The mismatch between the growing demand for psychological counseling and the limited availability of services has motivated research into the application of Large Language Models (LLMs) in this domain. Consequently, there is a need for a robust and unified benchmark to assess the counseling competence of various LLMs. Existing works, however, are limited by unprofessional client simulation, static question-and-answer evaluation formats, and unidimensional metrics. These limitations hinder their effectiveness in assessing a model's comprehensive ability to handle diverse and complex clients. To address this gap, we introduce CARE-Bench, a dynamic and interactive automated benchmark. It is built upon diverse client profiles derived from real-world counseling cases and simulated according to expert guidelines. CARE-Bench provides a multidimensional performance evaluation grounded in established psychological scales. Using CARE-Bench, we evaluate several general-purpose LLMs and specialized counseling models, revealing their current limitations. In collaboration with psychologists, we conduct a detailed analysis of the reasons for LLMs' failures when interacting with clients of different types, which provides directions for developing more comprehensive, universal, and effective counseling models.

Joint multilingual instruction tuning is a widely adopted approach to improve the multilingual instruction-following ability and downstream performance of large language models (LLMs), but the resulting multilingual capability remains highly sensitive to the composition and selection of the training data. Existing selection methods, often based on features like text quality, diversity, or task relevance, typically overlook the intrinsic linguistic structure of multilingual data. In this paper, we propose LangGPS, a lightweight two-stage pre-selection framework guided by language separability—a signal that quantifies how well samples in different languages can be distinguished in the model’s representation space. LangGPS first filters training data based on separability scores and then refines the subset using existing selection methods. Extensive experiments across six benchmarks and 22 languages demonstrate that applying LangGPS on top of existing selection methods improves their effectiveness and generalizability in multilingual training, especially for understanding tasks and low-resource languages. Further analysis reveals that highly separable samples facilitate the formation of clearer language boundaries and support faster adaptation, while low-separability samples tend to function as bridges for cross-lingual alignment. Besides, we also find that language separability can serves as an effective signal for multilingual curriculum learning, where interleaving samples with diverse separability levels yields stable and generalizable gains. Together, we hope our work offers a new perspective on data utility in multilingual contexts and support the development of more linguistically informed LLMs.

As large language models (LLMs) are increasingly deployed across culturally diverse regions, ensuring that their responses align with users’ cultural norms has become a critical challenge. Existing approaches to cultural alignment primarily rely on prompting or data-augmentation-based supervised finetuning, which teach models to follow norms indirectly through example-based supervision. However, these methods are difficult to scale and often fail to generalize, particularly in low-resource cultural settings. In this work, we propose CultureRL, a culture-norm-driven reinforcement learning framework that directly encodes cultural principles into model behavior. Rather than relying on output imitation, CultureRL provides normative feedback during training, enabling the model to internalize high-level cultural rules. It consists of two key components: (1) Norm Pool Construction (NPC), which clusters data from the World Values Survey into abstract cultural concepts to form a structured and retrievable norm pool; and (2) Norm Cluster-based Reward Mechanism (NCRM), which retrieves the relevant norm for each input and uses an external reward model to assess conformity, guiding model updates through cultural alignment. We evaluate CultureRL in both one-for-one (per-culture) and one-for-all (multi-culture) settings across nine cultures and three benchmarks. Results show that CultureRL consistently outperforms strong baselines, especially in terms of cultural consistency and adaptability.

Recent advancements in Large Reasoning Models (LRMs), such as OpenAI's o1/o3 and DeepSeek-R1, have demonstrated remarkable performance in specialized reasoning tasks through human-like deliberative thinking and long chain-of-thought reasoning. However, our systematic evaluation across various model families (DeepSeek, Qwen, and LLaMA) and scales (7B to 32B) reveals that acquiring these deliberative reasoning capabilities significantly reduces the foundational capabilities of LRMs, including notable declines in helpfulness and harmlessness, alongside substantially increased inference costs. Importantly, we demonstrate that adaptive reasoning---employing modes like Zero-Thinking, Less-Thinking, and Summary-Thinking---can effectively alleviate these drawbacks. Our empirical insights underline the critical need for developing more versatile LRMs capable of dynamically allocating inference-time compute according to specific task characteristics.

Ruihan Chen, Qiming Li, Xiaocheng Feng, Weihong Zhong, Xiaoliang Yang, Yuxuan Gu, Zekun Zhou, Yunfei Lu, Haoyu Ren, Kun Chen, Dandan Tu, Bing Qin. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Yanrui Du, Fenglei Fan, Sendong Zhao, Jiawei Cao, Ming Ma, Danyang Zhao, Shuren Qi, Ting Liu, Bing Qin. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Xiachong Feng, Deyi Yin, Xiaocheng Feng, Yi Jiang, Libo Qin, Yangfan Ye, Lei Huang, Weitao Ma, Qiming Li, Yuxuan Gu, Bing Qin, Lingpeng Kong. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Jiahe Guo, Xiangran Guo, Yulin Hu, Zimo Long, Xingyu Sui, Xuda Zhi, Yongbo Huang, Hao He, Weixiang Zhao, Yanyan Zhao, Bing Qin. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Retrieval-Augmented Generation (RAG) enhances Large Language Models (LLMs), especially for knowledge-intensive tasks.Despite its advantages, current RAG methods often struggle to fully exploit knowledge during generation.In particular, the synergy between the model's internal parametric knowledge and external retrieved knowledge remains limited.Retrieved contents may sometimes mislead generation, while certain generated content can guide the model toward more accurate outputs.In this work, we propose Collaborative Chainof-Agents, a framework designed to enhance explicitly synergy over both parametric and retrieved knowledge.Specifically, we first introduce CoCoA-zero, a multi-agent RAG framework that first performs conditional knowledge induction and then reasons answers.Building on this, we develop CoCoA, a long-chain training strategy that synthesizes extended multiagent reasoning trajectories from CoCoA-zero to fine-tune the LLM.This strategy enhances the model's capability to explicitly integrate and jointly leverage parametric and retrieved knowledge.Experimental results demonstrate the superiority of CoCoA in open-domain QA and multi-hop QA.Code is public 1 .

Qiming Li, Xiaocheng Feng, Yixuan Ma, Ruihan Chen, Zihe Tong, Zekai Ye, Xiachong Feng, Libo Qin, Haoyu Ren, Kun Chen, Yunfei Lu, Dandan Tu, Bing Qin. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Runxuan Liu, Xianhao Ou, Xinyan Ma, Jiyuan Wang, Jiafeng Liang, Jiaqi Li, Tao He, Zheng Chu, Rongchuan Mu, Zekun Wang, Baoxin Wang, Dayong Wu, Ming Liu, Shijin Wang, Guoping Hu, Bing Qin. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Weitao Ma, Xiaocheng Feng, Lei Huang, Xiachong Feng, Zhanyu Ma, Jun Xu, Jiuchong Gao, Jinghua Hao, Renqing He, Bing Qin. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Rongchuan Mu, Zexin Wang, Qianyu Wang, MingHua Ma, Zekun Wang, Ming Liu, Bing Qin. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Emotional Support Conversation requires not only affective expression but also grounded instrumental support to provide trustworthy guidance.However, existing ESC systems and benchmarks largely focus on affective support in text-only settings, overlooking how external tools can enable factual grounding and reduce hallucination in multi-turn emotional support.We introduce TEA-Bench, the first interactive benchmark for evaluating tool-augmented agents in ESC, featuring realistic emotional scenarios, an MCP-style tool environment, and process-level metrics that jointly assess the quality and factual grounding of emotional support.Experiments on nine LLMs show that tool augmentation generally improves emotional support quality and reduces hallucination, but the gains are strongly capacity-dependent: stronger models use tools more selectively and effectively, while weaker models benefit only marginally.We further release TEA-Dialog, a dataset of toolenhanced ESC dialogues, and find that supervised fine-tuning improves in-distribution support but generalizes poorly.Our results underscore the importance of tool use in building reliable emotional support agents. 1

Boyu Xiao, Xiuqi Tian, Xuwen Song, Haochun Wang, Guanchun Song, Sendong Zhao, Bing Qin. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Zekun Yuan, Yangfan Ye, Xiaocheng Feng, Baohang Li, Qichen Hong, Yunfei Lu, Dandan Tu, Bing Qin. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Liang Zhao, Xiaocheng Feng, Weihong Zhong, Lei Huang, Kun Zhu, Baoxin Wang, Dayong Wu, Guoping Hu, Ting Liu, Bing Qin. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Yang Zhao, Yangou Ouyang, Xiao Ding, Hepeng Wang, Bibo Cai, Kai Xiong, Jinglong Gao, Zhouhao Sun, Li Du, Bing Qin, Ting Liu. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Yang Zhao, Hepeng Wang, Xiao Ding, Yangou Ouyang, Bibo Cai, Kai Xiong, Jinglong Gao, Zhouhao Sun, Li Du, Bing Qin, Ting Liu. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Generative Recommenders (GRs), exemplified by the Hierarchical Sequential Transduction Unit (HSTU), have emerged as a powerful paradigm for modeling long user interaction sequences. However, we observe that their ''flat-sequence'' assumption overlooks the rich, intrinsic structure of user behavior. This leads to two key limitations: a failure to capture the temporal hierarchy of session-based engagement, and computational inefficiency, as dense attention introduces significant noise that obscures true preference signals within semantically sparse histories, which deteriorates the quality of the learned representations. To this end, we propose a novel framework named HPGR (Hierarchical and Preference-aware Generative Recommender), built upon a two-stage paradigm that injects these crucial structural priors into the model to handle the drawback. Specifically, HPGR comprises two synergistic stages. First, a structure-aware pre-training stage employs a session-based Masked Item Modeling (MIM) objective to learn a hierarchically-informed and semantically rich item representation space. Second, a preference-aware fine-tuning stage leverages these powerful representations to implement a Preference-Guided Sparse Attention mechanism, which dynamically constrains computation to only the most relevant historical items, enhancing both efficiency and signal-to-noise ratio. Empirical experiments on a large-scale proprietary industrial dataset from APPGallery and an online A/B test verify that HPGR achieves state-of-the-art performance over multiple strong baselines, including HSTU and MTGR.

Abstract Machine Unlearning (MU) has emerged as a promising technique for aligning large language models (LLMs) with safety requirements to steer them forgetting specific harmful contents. Despite the significant progress in previous studies, we argue that the current evaluation criteria, which solely focus on safety evaluation, are actually impractical and biased , leading to concerns about the true effectiveness of MU techniques. To address this, we propose to comprehensively evaluate LLMs after MU from three aspects: safety, over-safety, and general utility. Specifically, a novel benchmark M u B ench with 18 related datasets is first constructed, where the safety is measured with both vanilla harmful inputs and 10 types of jailbreak attacks. Furthermore, we examine whether MU introduces side effects, focusing on over-safety and utility-loss. Extensive experiments are performed on 3 popular LLMs with 7 recent MU methods. The results highlight a challenging trilemma in safety alignment without side effects, indicating that there is still considerable room for further exploration. M u B ench serves as a comprehensive benchmark, fostering future research on MU for safety alignment of LLMs.

Abstract Large Language Models (LLMs) exhibit unique strengths arising from differences in model architecture, training data, and strategies. Ensemble learning has been explored to leverage these complementary strengths through decision-level sharing (i.e.,Decision Ensemble), which combines the predictions from multiple LLMs. However, such methods integrate only shallow decisions and overlook the exchange of deeper levels of information within the internal representations of LLMs, such as problem understanding, world knowledge, and latent reasoning patterns. In this work, we propose Representation Ensemble (RISE), a novel ensemble framework that enables cross-LLM representation sharing for richer information exchange. To address challenges of representation-level interaction caused by layer misalignment and latent-space incompatibility across LLMs, we introduce a representation alignment method based on relational similarity measures and an orthogonal latent-space transformation. Experimental results show that (1) RISE achieves performance competitive with existing decision ensemble methods, and (2) RISE is strongly complementary to decision ensemble, with their combination boosting collaboration gains by 14%–41%. Finally, we further compare ensemble of small LLMs to a single larger LLM and to model merging and composition approaches, and find that ensemble learning consistently generalizes well without additional training.

Video question answering (VideoQA) necessitates simultaneous understanding of visual and linguistic information, requiring both in-depth analysis of individual modality features and the establishment of cross-modal correlations to achieve precise reasoning. However, VideoQA models often struggle with irrelevant temporal and spatial noise due to the dense events and concepts in real-world complex video contents. Previous works reduce noise by only sampling a fixed number of visual tokens at the patch level, overlooking the variation in the required granularities of features and quantities of visual cues across different question conditions. To address these, we propose an Aggregating-then-Pruning Sampler (APSam), which diversifies feature granularities and adaptively denoises on a per-question basis. Specifically, we propose a conditional token aggregator to obtain multi-granularity visual semantics by merging similar question-relevant tokens. Then, we propose a conditional token pruner, which restricts noise tokens through a variable-capacity receptive field determined by the inputs. Experimental results show that APSam achieves significant performance on three challenging complex VideoQA datasets,i.e., AGQAv2, NExT-QA, and STAR. Further analyses reveal that the APSam also exhibits high reasoning capability and interpretability.

Multi-hop Knowledge Graph Reasoning (KGR) seeks to identify accurate answers within Knowledge Graphs (KGs) via multi-step reasoning, predominantly utilizing reinforcement learning (RL) to enhance the efficiency of the reasoning process. Unlike traditional Knowledge Graph Embedding (KGE) methods, RL-based approaches offer superior interpretability. However, these methods often underperform due to two critical limitations: (1) their over-reliance on Horn rules for reasoning paths, which restricts their expressive power; and (2) inadequate utilization of reasoning states during the process. To address these issues, we propose a novel RL-based framework, RAR, which shifts focus from individual paths to subgraph structures for more robust predictions. RAR frames the retrieval of reasoning subgraphs from the KG as a Markov Decision Process (MDP) and incorporates a subgraph retriever. To efficiently explore the extensive subgraph space, we integrate multi-agent RL to enhance the retriever's capabilities. Additionally, RAR features an advanced analyst module that meticulously examines reasoning states. These modules function iteratively: the retriever expands the subgraph, followed by the analyst module's in-depth analysis. The insights gained are then used to inform subsequent retrieval steps. Ultimately, the predicted scores from both modules are synthesized to produce more precise posterior scores. Experimental results across multiple datasets demonstrate RAR's efficacy, showcasing a notable improvement over existing state-of-the-art RL-based KGR methods.

Large Language Models (LLMs) have gained significant attention for their exceptional performance across various domains. Despite their advancements, concerns persist regarding their implicit bias, which often leads to negative social impacts. Therefore, it is essential to identify the implicit bias in LLMs and investigate the potential threat posed by it. Our study focused on a specific type of implicit bias, termed the ''Yes-No'' implicit bias, which refers to LLMs' inherent tendency to favor ''Yes'' or ''No'' responses to a single instruction. By comparing the probability of LLMs generating a series of ''Yes'' versus ''No'' responses, we observed different inherent response tendencies exhibited by LLMs when faced with different instructions. To further investigate the impact of such bias, we developed an attack method called Implicit Bias In-Context Manipulation, attempting to manipulate LLMs' behavior. Specifically, we explored whether the ''Yes'' implicit bias could manipulate ''No'' responses into ''Yes'' in LLMs' responses to malicious instructions, leading to harmful outputs. Our findings revealed that the ''Yes'' implicit bias brings a significant security threat, comparable to that of carefully designed attack methods. Moreover, we offered a comprehensive analysis from multiple perspectives to deepen the understanding of this security threat, emphasizing the need for ongoing improvement in LLMs' security.

Recent advancements in proactive dialogues have garnered significant attention, particularly for more complex objectives (e.g. emotion support and persuasion). Unlike traditional task-oriented dialogues, proactive dialogues demand advanced policy planning and adaptability, requiring rich scenarios and comprehensive policy repositories to develop such systems. However, existing approaches tend to rely on Large Language Models (LLMs) for user simulation and online learning, leading to biases that diverge from realistic scenarios and result in suboptimal efficiency. Moreover, these methods depend on manually defined, context-independent, coarse-grained policies, which not only incur high expert costs but also raise concerns regarding their completeness. In our work, we highlight the potential for automatically discovering policies directly from raw, real-world dialogue records. To this end, we introduce a novel dialogue policy planning framework, LDPP. It fully automates the process from mining policies in dialogue records to learning policy planning. Specifically, we employ a variant of the Variational Autoencoder to discover fine-grained policies represented as latent vectors. After automatically annotating the data with these latent policy labels, we propose an Offline Hierarchical Reinforcement Learning (RL) algorithm in the latent space to develop effective policy planning capabilities. Our experiments demonstrate that LDPP outperforms existing methods on two proactive scenarios, even surpassing ChatGPT with only a 1.8-billion-parameter LLM.

Large language models (LLMs) have demonstrated significant progress in multilingual language understanding and generation. However, due to the imbalance in training data, their capabilities in non-English languages are limited. Recent studies revealed the English-pivot multilingual mechanism of LLMs, where LLMs implicitly convert non-English queries into English ones at the bottom layers and adopt English for thinking at the middle layers. However, due to the absence of explicit supervision for cross-lingual alignment in the intermediate layers of LLMs, the internal representations during these stages may become inaccurate. In this work, we introduce a deep supervision fine-tuning method (DFT) that incorporates additional supervision in the internal layers of the model to guide its workflow. Specifically, we introduce two training objectives on different layers of LLMs: one at the bottom layers to constrain the conversion of the target language into English, and another at the middle layers to constrain reasoning in English. To effectively achieve the guiding purpose, we designed two types of supervision signals: logits and feature, which represent a stricter constraint and a relatively more relaxed guidance. Our method guides the model to not only consider the final generated result when processing non-English inputs but also ensure the accuracy of internal representations. We conducted extensive experiments on typical English-centric large models, LLaMA-2 and Gemma-2, and the results on multiple multilingual datasets show that our method significantly outperforms traditional fine-tuning methods.

Recent Large Vision-Language Models (LVLMs) have shown promising reasoning capabilities on text-rich images from charts, tables, and documents. However, the abundant text within such images may increase the model's sensitivity to language. This raises the need to evaluate LVLM performance on cross-lingual text-rich visual inputs, where the language in the image differs from the language of the instructions. To address this, we introduce XT-VQA (Cross-Lingual Text-Rich Visual Question Answering), a benchmark designed to assess how LVLMs handle language inconsistency between image text and questions. XT-VQA integrates five existing text-rich VQA datasets and a newly collected dataset, XPaperQA, covering diverse scenarios that require faithful recognition and comprehension of visual information despite language inconsistency. Our evaluation of prominent LVLMs on XT-VQA reveals a significant drop in performance for cross-lingual scenarios, even for models with multilingual capabilities. A mutual information analysis suggests that this performance gap stems from cross-lingual questions failing to adequately activate relevant visual information. To mitigate this issue, we propose MVCL-MI (Maximization of Vision-Language Cross-Lingual Mutual Information), where a visual-text cross-lingual alignment is built by maximizing mutual information between the model's outputs and visual information. This is achieved by distilling knowledge from monolingual to cross-lingual settings through KL divergence minimization, where monolingual output logits serve as a teacher. Experimental results on the XT-VQA demonstrate that MVCL-MI effectively reduces the visual-text cross-lingual performance disparity while preserving the inherent capabilities of LVLMs, shedding new light on the potential practice for improving LVLMs.

Zekun Wang, MingHua Ma, Zexin Wang, Rongchuan Mu, Liping Shan, Ming Liu, Bing Qin. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

Large language models (LLMs) have mastered abundant simple and explicit commonsense knowledge through pre-training, enabling them to achieve human-like performance in simple commonsense reasoning. Nevertheless, LLMs struggle to reason with complex and implicit commonsense knowledge that is derived from simple ones (such as understanding the long-term effects of certain events), an aspect humans tend to focus on more. Existing works focus on complex tasks like math and code, while complex commonsense reasoning remains underexplored due to its uncertainty and lack of structure. To fill this gap and align with real-world concerns, we propose a benchmark Com^2 focusing on complex commonsense reasoning. We first incorporate causal event graphs to serve as structured complex commonsense. Then we adopt causal theory (e.g., intervention) to modify the causal event graphs and obtain different scenarios that meet human concerns. Finally, an LLM is employed to synthesize examples with slow thinking, which is guided by the logical relationships in the modified causal graphs. Furthermore, we use detective stories to construct a more challenging subset. Experiments show that LLMs struggle in reasoning depth and breadth, while post-training and slow thinking can alleviate this. The code and data are available at https://github.com/Waste-Wood/Com2.

Yexing Du, Youcheng Pan, Ziyang Ma, Bo Yang, Yifan Yang, Keqi Deng, Xie Chen, Yang Xiang, Ming Liu, Bing Qin. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

Recent studies provide large language models (LLMs) with textual task-solving experiences via prompts to improve their performance.However, previous methods rely on substantial human labor or time to gather such experiences for each task, which is impractical given the growing variety of task types in user queries to LLMs.To address this issue, we design an autonomous experience transfer framework to explore whether LLMs can mimic human cognitive intelligence to autonomously transfer experience from existing source tasks to newly encountered target tasks. This not only allows the acquisition of experience without extensive costs of previous methods, but also offers a novel path for the generalization of LLMs.Experimental results on 13 datasets demonstrate that our framework effectively improves the performance of LLMs. Furthermore, we provide a detailed analysis of each module in the framework.

Yuxuan Gu, Wenjie Wang, Xiaocheng Feng, Weihong Zhong, Kun Zhu, Lei Huang, Ting Liu, Bing Qin, Tat-Seng Chua. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

Lei Huang, Xiaocheng Feng, Weitao Ma, Yuchun Fan, Xiachong Feng, Yuxuan Gu, Yangfan Ye, Liang Zhao, Weihong Zhong, Baoxin Wang, Dayong Wu, Guoping Hu, Lingpeng Kong, Tong Xiao, Ting Liu, Bing Qin. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

Lei Huang, Xiaocheng Feng, Weitao Ma, Yuchun Fan, Xiachong Feng, Yangfan Ye, Weihong Zhong, Yuxuan Gu, Baoxin Wang, Dayong Wu, Guoping Hu, Bing Qin. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

The Retrieval-Augmented Generation (RAG) framework introduces a retrieval module to dynamically inject retrieved information into the input context of large language models (LLMs), and has demonstrated significant success in various NLP tasks.However, the current study points out that there is a preference gap between retrievers and LLMs in the RAG framework, which limit the further improvement of system performance.Some highly relevant passages may interfere with LLM reasoning because they contain complex or contradictory information; while some indirectly related or even inaccurate content may help LLM generate more accurate answers by providing suggestive information or logical clues.To solve this, we propose GainRAG, a novel approach that aligns the retriever's and LLM's preferences by defining a new metric, "gain", which measure how well an input passage contributes to correct outputs.Specifically, we propose a method to estimate these gain signals and train a middleware that aligns the preferences of the retriever and the LLM using only limited data.In addition, we introduce a pseudo-passage strategy to mitigate degradation.The experimental results on 6 datasets verify the effectiveness of GainRAG 1 .

Jiafeng Liang, Shixin Jiang, Xuan Dong, Ning Wang, Zheng Chu, Hui Su, Jinlan Fu, Ming Liu, See-Kiong Ng, Bing Qin. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

Runxuan Liu, Luobei Luobei, Jiaqi Li, Baoxin Wang, Ming Liu, Dayong Wu, Shijin Wang, Bing Qin. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

Weitao Ma, Xiyuan Du, Xiaocheng Feng, Lei Huang, Yichong Huang, Huiyi Zhang, Xiaoliang Yang, Baohang Li, Xiachong Feng, Ting Liu, Bing Qin. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

Guo Tang, Zheng Chu, Wenxiang Zheng, Junjia Xiang, Yizhuo Li, Weihao Zhang, Ming Liu, Bing Qin. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

Multi-agent collaboration has emerged as a pivotal paradigm for addressing complex, distributed tasks in large language model (LLM)-driven applications. While prior research has focused on high-level architectural frameworks, the granular mechanisms governing agents—critical to performance and scalability—remain underexplored. This study systematically investigates four dimensions of collaboration strategies: (1) agent governance, (2) participation control, (3) interaction dynamics, and (4) dialogue history management. Through rigorous experimentation under two context-dependent scenarios—Distributed Evidence Integration (DEI) and Structured Evidence Synthesis (SES)—we quantify the impact of these strategies on both task accuracy and computational efficiency. Our findings reveal that centralized governance, instructor-led participation, ordered interaction patterns, and instructor-curated context summarization collectively optimize the trade-off between decision quality and resource utilization with the support of the proposed Token-Accuracy Ratio (TAR). This work establishes a foundation for designing adaptive, scalable multi-agent systems, shifting the focus from structural novelty to strategic interaction mechanics.

Yangfan Ye, Xiaocheng Feng, Zekun Yuan, Xiachong Feng, Libo Qin, Lei Huang, Weitao Ma, Yichong Huang, Zhirui Zhang, Yunfei Lu, Xiaohui Yan, Duyu Tang, Dandan Tu, Bing Qin. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

Large Vision-Language Models (LVLMs) have demonstrated impressive multimodal abilities but remain prone to multilingual object hallucination, with a higher likelihood of generating responses inconsistent with the visual input when utilizing queries in non-English languages compared to English. Most existing approaches to address these rely on pretraining or fine-tuning, which are resource-intensive. In this paper, inspired by observing the disparities in cross-modal attention patterns across languages, we propose Cross-Lingual Attention Intervention for Mitigating multilingual object hallucination (CLAIM) in LVLMs, a novel near training-free method by aligning attention patterns. CLAIM first identifies language-specific cross-modal attention heads, then estimates language shift vectors from English to the target language, and finally intervenes in the attention outputs during inference to facilitate cross-lingual visual perception capability alignment. Extensive experiments demonstrate that CLAIM achieves an average improvement of 13.56% (up to 30% in Spanish) on the POPE and 21.75% on the hallucination subsets of the MME benchmark across various languages. Further analysis reveals that multilingual attention divergence is most prominent in intermediate layers, highlighting their critical role in multilingual scenarios.

LLMs’ performance on complex tasks is still unsatisfactory. A key issue is that presently LLMs learn in a data-driven schema, while the instructions about these complex tasks are both scarce and hard to collect or construct. On the contrary, a prominent phenomenon is that LLMs can learn rather fast on simpler tasks with adequate prior knowledge captured during pretraining stage. Thus, if the prerequisite and mechanism of such rapid generalization could be elucidated, it could enhance the efficiency and effectiveness of the LLM’s ability to learn complex tasks. Thus, in this paper, we employ a gradient-based method, to dissect the process that the SFT process adapts LLMs to downstream tasks via the perspective of attention patterns. We find that: (1) LLMs selectively activate task-specific attention heads during SFT; (2) activation patterns for complex tasks are combinations of basic task patterns; and (3) changes in a few parameters can significantly impact activation patterns after SFT on a small number of samples.Based on these insights, experiments are conducted to actually enhance the efficiency and effectiveness of SFT.

Yang Zhao, Li Du, Xiao Ding, Yangou Ouyang, Hepeng Wang, Kai Xiong, Jinglong Gao, Zhouhao Sun, Dongliang Xu, Qing Yang, Dongchen Li, Bing Qin, Ting Liu. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

Weixiang Zhao, Yulin Hu, Yang Deng, Jiahe Guo, Xingyu Sui, Xinyang Han, An Zhang, Yanyan Zhao, Bing Qin, Tat-Seng Chua, Ting Liu. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

Weixiang Zhao, Yulin Hu, Yang Deng, Tongtong Wu, Wenxuan Zhang, Jiahe Guo, An Zhang, Yanyan Zhao, Bing Qin, Tat-Seng Chua, Ting Liu. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

The colossal parameters and computational overhead of Large Language Models (LLMs) challenge their real-world applications. Network pruning, which targets unstructured or structured sparsity by removing redundant parameters, has recently been explored for LLM acceleration. Existing LLM pruning works focus on unstructured pruning, which typically requires special hardware support for a practical speed-up. In contrast, structured pruning can reduce latency on general devices. However, it remains a challenge to perform structured pruning efficiently and maintain performance, especially at high sparsity ratios. To this end, we introduce an efficient structured pruning framework named CFSP, which leverages both Coarse (interblock) and Fine-grained (intrablock) activation information as an importance criterion to guide pruning. The pruning is highly efficient, as it only requires one forward pass to compute feature activations. Specifically, we first allocate the sparsity budget across blocks based on their importance and then retain important weights within each block. In addition, we introduce a recovery fine-tuning strategy that adaptively allocates training overhead based on coarse-grained importance to further improve performance. Experimental results demonstrate that CFSP outperforms existing methods on diverse models across various sparsity budgets. Our code will be available at https://github.com/wyxscir/CFSP.

Despite the remarkable reasoning capabilities demonstrated by large language models (LLM), the substantial computational overhead limits their practices. Some efforts have been directed toward distilling multi-step reasoning capabilities into smaller models through chain-of-thought (CoT). While CoT facilitates multi-step reasoning, the dependencies between reasoning steps are not always clearly discernible, which may lead to inconsistent reasoning. In this paper, we introduce fine-grained attribution reasoning distillation (FARD), which incorporates grounded citations to consolidate the relationships between reasoning steps. Specifically, FARD distills attribution reasoning rationales from LLMs to substitute CoT reasonings, which clarifies the dependencies among reasoning steps. Besides, we regularize the model’s attention pattern by leveraging the causal dependencies between reasoning steps, thereby enhancing the consistency of reasoning. Grounded attribution reasoning also enhances interpretability and verifiability, thereby facilitating faithful reasoning. We evaluate FARD on mathematical and general reasoning benchmarks. The experimental results indicate that FARD outperforms CoT distillation methods in mathematical reasoning, demonstrating its effectiveness. Furthermore, the small models trained with FARD have shown outstanding performance in out-of-distribution reasoning, proving strong generalization capabilities.

Large language model unlearning has garnered increasing attention due to its potential to address security and privacy concerns, leading to extensive research in the field. However, existing studies have predominantly focused on instance-level unlearning, specifically targeting the removal of predefined instances containing sensitive content. This focus has left a gap in the exploration of removing an entire entity, which is critical in real-world scenarios such as copyright protection. To close this gap, we propose a novel task named Entity-level unlearning, which aims to erase entity-related knowledge from the target model completely. To investigate this task, we systematically evaluate popular unlearning algorithms, revealing that current methods struggle to achieve effective entity-level unlearning. Then, we further explore the factors that influence the performance of unlearning algorithms, identifying that the knowledge coverage of the forget set and its size play pivotal roles. Notably, our analysis also uncovers that entities introduced through fine-tuning are more vulnerable than pre-trained entities during unlearning. We hope these findings can inspire future improvements in entity-level unlearning for LLMs.

In the field of NLP, Large Language Models (LLMs) have markedly enhanced performance across a variety of tasks. However, the comprehensive evaluation of LLMs remains an inevitable challenge for the community. Recently, the adoption of Multiple Choice Question Answering (MCQA) as a benchmark for assessing LLMs has gained considerable traction. However, concerns regarding the robustness of this evaluative method persist. Building upon previous discussions on the issue of variability, we reveal an additional dimension of concern: LLMs may perform MCQA by selecting the least incorrect option rather than distinctly correct. This observation suggests that LLMs might regard multiple options as correct, which could undermine the reliability of MCQA as a metric for evaluating LLMs. To address this challenge, we introduce an enhanced dataset augmentation method for MCQA, termed MCQA+, to provide a more accurate reflection of the performance, thereby highlighting the necessity for more sophisticated evaluation mechanisms in the assessment of LLM capabilities.

Evaluating the graph comprehension and reasoning abilities of Large Language Models (LLMs) is challenging and often incomplete. Existing benchmarks focus primarily on pure graph understanding, lacking a comprehensive evaluation across all graph types and detailed capability definitions. This paper presents GraCoRe, a benchmark for systematically assessing LLMs’ graph comprehension and reasoning. GraCoRe uses a three-tier hierarchical taxonomy to categorize and test models on pure graph and heterogeneous graphs, subdividing capabilities into 10 distinct areas tested through 19 tasks. Our benchmark includes 11 datasets with 5,140 graphs of varying complexity. We evaluate four closed-source and eight open-source LLMs, conducting thorough analyses from both ability and task perspectives. Key findings reveal that OpenAI o1 model has amazing comprehension and reasoning capabilities, semantic enrichment enhances reasoning performance, node ordering impacts task success, and the ability to process longer texts does not necessarily improve graph comprehension or reasoning.

Entity alignment (EA), critical for knowledge graph (KG) integration, identifies equivalent entities across different KGs.Traditional methods often face challenges in semantic understanding and scalability.The rise of language models (LMs), particularly large language models (LLMs), has provided powerful new strategies.This paper systematically reviews LM-driven EA methods, proposing a novel taxonomy that categorizes methods in three key stages: data preparation, feature embedding, and alignment.We further summarize key benchmarks, evaluation metrics, and discuss future directions.This paper aims to provide researchers and practitioners with a clear and comprehensive understanding of how language models reshape the field of entity alignment.* These authors contributed equally.

In recent years, Large Language Models (LLMs) have made significant progress in emotional support dialogue.However, there are two major challenges for LLM-based support systems.First, users may be hesitant to fully disclose their emotions at the outset.Second, direct probing or excessive questioning can induce discomfort or even resistance.To bridge this gap, we propose COCOON, a proactive emotional support framework that leverages principles of active listening to uncover implicit user needs.We design a multi-stage data curation pipeline and an annotation mechanism for support strategies.Based on this framework, we build COCOON-Llama3, a fine-tuned large language model, and evaluate it using both standard metrics and psychological scales.Experimental results indicate that our model more effectively elicits implicit emotional needs and delivers empathetic support compared to existing baselines, suggesting its utility for building more inclusive emotional support dialogue systems.

Detecting depression through users’ social media posting history is crucial for enabling timely intervention; however, irrelevant content within these posts negatively impacts detection performance. Thus, it is crucial to extract pertinent content from users’ complex posting history. Current methods utilize frozen screening models, which can miss critical information and limit overall performance due to isolated screening and detection processes. To address these limitations, we propose E2-LPS End-to-End Learnable Psychiatric Scale Guided Risky Post Screening Model) for jointly training our screening model, guided by psychiatric scales, alongside the detection model. We employ a straight-through estimator to enable a learnable end-to-end screening process and avoid the non-differentiability of the screening process. Experimental results show that E2-LPS outperforms several strong baseline methods, and qualitative analysis confirms that it better captures users’ mental states than others.

Graph-theoretic problems arise in real-world applications like logistics, communication networks, and traffic optimization.These problems are often complex, noisy, and irregular, posing challenges for traditional algorithms.Large language models offer potential solutions but face several challenges, including limited accuracy, input length constraints, and suboptimal algorithm selection.To address these challenges, we propose MA-GTS (Multi-Agent Graph Theory Solver), a multi-agent framework that decomposes these complex problems through agent collaboration.MA-GTS maps the implicitly expressed textbased graph data into clear, structured graph representations and dynamically selects the most suitable algorithm based on problem constraints and graph structure scale.We validate MA-GTS using the G-REAL dataset, a real-world-inspired graph theory dataset we created.Experimental results show that MA-GTS outperforms state-of-the-art methods in cost-effectiveness, accuracy, and scalability, achieving strong results on multiple benchmarks (G-REAL 93.6%, GraCoRe 96.9% NL-Graph 98.4%) with robust performance on both closed-and open-source base models.

Yirong Zeng, Xiao Ding, Yuxian Wang, Weiwen Liu, Yutai Hou, Wu Ning, Xu Huang, Duyu Tang, Dandan Tu, Bing Qin, Ting Liu. Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing. 2025.

Weixiang Zhao, Jiahe Guo, Yulin Hu, Yang Deng, An Zhang, Xingyu Sui, Xinyang Han, Yanyan Zhao, Bing Qin, Tat-Seng Chua, Ting Liu. Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing. 2025.

Understanding the internal origins of capabilities in large language models (LLMs) is crucial for interpretability and efficient adaptation.However, the emergence of specific capabilities remains poorly understood, as most existing approaches rely on external signals (e.g., performance shifts or gradient similarities) with limited structural grounding.To address these issues, this paper proposes a lightweight and highly interpretable approach that links LLM capabilities to internal components by identifying correspondences at the level of attention heads.Specifically, we first define five fundamental capabilities, namely Mathematical Reasoning, Reading Comprehension, Commonsense Reasoning, Scientific Reasoning, and Professional Expertise, and employ probing techniques to detect the attention heads most predictive of each, thereby establishing capability-head mappings.For targeted instruction tuning, complex tasks are decomposed into these fundamental capabilities, and training data are selected accordingly.Experiments on LLaMA3.1-8B and Qwen2.5-7Bshow over 70% discrimination accuracy in identifying capabilities.On MMLU and BBH, our method improves accuracy by 1 to 1.5 points over the gradient-based method LESS and by 5 to 6 points over other intermediate-state baselines 1 .

The rapid growth of scientific literature demands efficient methods to organize and synthesize research findings.Existing taxonomy construction methods, leveraging unsupervised clustering or direct prompting of large language models (LLMs), often lack coherence and granularity.We propose a novel context-aware hierarchical taxonomy generation framework that integrates LLM-guided multi-aspect encoding with dynamic clustering.Our method leverages LLMs to identify key aspects of each paper (e.g., methodology, dataset, evaluation) and generates aspect-specific paper summaries, which are then encoded and clustered along each aspect to form a coherent hierarchy.In addition, we introduce a new benchmark of 156 expert-crafted taxonomies encompassing 11.6 k papers, providing the first naturally annotated dataset for this task.Experimental results demonstrate that our method significantly outperforms prior approaches, achieving stateof-the-art performance in taxonomy coherence, granularity, and interpretability. 1

Large Language Models (LLMs), such as the GPT series, have achieved remarkable performance in question answering through large-scale pretraining. However, LLMs often lack transparency in their reasoning processes and struggle with hallucination. To overcome these challenges, we propose Dual-NeSy, a Dual-system framework that integrates Neural networks with Symbolic logic for interpretable question answering. Specifically, Dual-NeSy leverages fast, heuristic reasoning for knowledge composition (System 1) and utilizes structured, multi-step symbolic reasoning (System 2) for more deliberate, logical verification. This combined approach enhances both the interpretability and accuracy of reasoning. Our approach outperforms previous methods in commonsense reasoning and reading comprehension tasks, achieving state-of-the-art results on three benchmark datasets: QASC, WorldTree, and WikiHop.

Diffusion Probabilistic Models (DPMs) have achieved significant success in generative tasks. However, their training and sampling processes suffer from the issue of distribution mismatch. During the denoising process, the input data distributions differ between the training and inference stages, potentially leading to inaccurate data generation. To obviate this, we analyze the training objective of DPMs and theoretically demonstrate that this mismatch can be alleviated through Distributionally Robust Optimization (DRO), which is equivalent to performing robustness-driven Adversarial Training (AT) on DPMs. Furthermore, for the recently proposed Consistency Model (CM), which distills the inference process of the DPM, we prove that its training objective also encounters the mismatch issue. Fortunately, this issue can be mitigated by AT as well. Based on these insights, we propose to conduct efficient AT on both DPM and CM. Finally, extensive empirical studies validate the effectiveness of AT in diffusion-based models. The code is available at https://github.com/kugwzk/AT_Diff.

In Emotion Recognition in Conversations (ERC), one-hot labels are typically used as ground truth, but they may not fully capture all emotions conveyed in an utterance. Recent work in textual ERC has investigated self-distillation techniques for generating soft labels via single-instance generation, aiming to improve emotional understanding. However, these approaches still struggle to fully capture complex emotional expressions. In multimodal ERC (MERC), generating soft labels is even more challenging due to integrating multiple modalities, which may express distinct emotions. Based on this, we propose a Multi-stage Multi-target Knowledge Distillation Framework, consisting of two components: Multi-stage Distillation (MSD) and Multi-target Distillation (MTD). MSD focuses on multi-stage self-distillation of soft labels and utterance representations, encouraging the MERC model to refine its label predictions across stages. Building on MSD, MTD further distills soft labels and features from a feature extractor used to extract modality-common and modality-specific features, deepening the model’s understanding of emotions in multimodal scenarios. Experimental results on two datasets show that our framework significantly improves the performance of various MERC models, surpassing state-of-the-art methods.

Large Language Models (LLMs) have demonstrated exceptional performance across various tasks, with pre-training stage serving as the cornerstone of their capabilities. However, the conventional fixed-length data composition strategy for pre-training presents several practical challenges. When using shorter sequences, documents are often truncated, potentially leading to information loss and affecting the model's ability to capture long-range dependencies. Conversely, longer sequences require concatenation of multiple documents, which can introduce noise and affect the natural document boundaries and semantic coherence as well as require substantial computational overhead. To address these challenges, we first establish three quantitative metrics for evaluating data composition quality: padding ratio, truncation ratio, and concatenation ratio. Building upon these metrics, we propose a novel multi-bucket data composition method that transcends the fixed-length paradigm. Our approach adaptively organizes training data to achieve optimal composition quality as measured by the proposed metrics, offering a more flexible and efficient approach for pre-training. We conduct extensive experiments and the results demonstrate that our proposed method significantly enhances both the efficiency and effectiveness of LLM pre-training. Our proposed method has been adopted in the Du Xiaoman–XuanYuan series of financial large language models at https://github.com/Duxiaoman-DI/XuanYuan.

Artificial intelligence advances drug design by predicting drug properties through the encoding of drug molecules. Since different molecular representations contain complementary information, a large amount of research is currently dedicated to the fusion of these representations. However, current multimodal molecular studies rely more on single-granularity methods, resulting in the loss of atomic-level information between representations. Inspired by the success of multi-granularity approaches, we introduce MolFusion, an innovative method for multi-granularity molecular representation fusion. MolFusion comprises two core components: MolSim for molecular-level fusion and AtomAlign for atomic-level fusion. Comprehensive experiments show that our method outperforms all powerful baselines in terms of average performance across both classification and regression tasks, with a particular enhancement in regression tasks. Our code and dataset will be available on https://github.com/Mengqi97/MolFusion.

As an increasing number of users share their lives and mental states on social media, many studies attempt to detect depression risk through social media videos using non-verbal cues like facial expressions, posture, gaze, and intonation from individual social media platforms, a proven effective field. However, these studies have focused on single-video level analysis to detect depression. These researches fail to capture the dynamic nature of social media streams and the complex, often gradual manifestation of depression. This limitation overlooks the comprehensive mental state of users, which can only be understood through their extended video histories. To address this, we introduce the Multimodal User-level Depression Detection Dataset (MUD3). MUD3 includes the long-term video histories of depressed users on social media platforms, containing user mental states across multiple videos and treating the video histories as a continuous social media stream. This allows us to model multiple videos at the user-level and analyze users' long-term mental states. MUD3 and supplementary materials are available at https://github.com/Syx1030/MUD3.

The integration of multimodal information, particularly visual content, into dialogue systems has primarily focused on interpreting user-provided inputs, while comparatively little attention has been given to the proactive use of such content to enhance responses. In this paper, we explore a new research direction that addresses this gap by enabling dialogue systems to autonomously determine when and how to supplement textual responses with relevant images, based on conversational context and user intent. To support this goal, we propose AMI (Automated Multimodal Insertion), a novel framework for dynamic, context-aware multimodal supplementation in dialogue. We also introduce RID (Response with Appropriate Image Dataset), a bilingual (Chinese-English) multimodal multi-turn dialogue dataset designed to train and evaluate systems on this capability. RID features fine-grained annotations on image insertion timing and rationale, along with carefully aligned image-text pairs to ensure semantic coherence. Our experiments demonstrate that models trained with RID not only generate more informative and engaging responses, but also exhibit a stronger ability to leverage visual content when it is truly beneficial. These findings highlight the potential of proactive multimodal supplementation and offer new insights for advancing the development of intelligent, human-like dialogue systems. Code and data are available at: https://github.com/Tanthen/SCIR-AMI.

Bichen Wang, Yuzhe Zi, Yixin Sun, Yanyan Zhao, Bing Qin. Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers). 2025.

Yirong Zeng, Xiao Ding, Bibo Cai, Ting Liu, Bing Qin. Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers). 2025.

Pre-trained language models represented by the Transformer have been proven to possess strong base capabilities, and the representative self-attention mechanism in the Transformer has become a classic in sequence modeling architectures. Different from the work of proposing sequence modeling architecture to improve the efficiency of attention mechanism, this work focuses on the impact of sequence modeling architectures on base capabilities. Specifically, our concern is: How exactly do sequence modeling architectures affect the base capabilities of pre-trained language models? In this work, we first point out that the mixed domain pre-training setting commonly adopted in existing architecture design works fails to adequately reveal the differences in base capabilities among various architectures. To address this, we propose a limited domain pre-training setting with out-of-distribution testing, which successfully uncovers significant differences in base capabilities among architectures at an early stage. Next, we analyze the base capabilities of stateful sequence modeling architectures, and find that they exhibit significant degradation in base capabilities compared to the Transformer. Then, through a series of architecture component analysis, we summarize a key architecture design principle: A sequence modeling architecture need possess full-sequence arbitrary selection capability to avoid degradation in base capabilities. Finally, we empirically validate this principle using an extremely simple Top-1 element selection architecture and further generalize it to a more practical Top-1 chunk selection architecture. Experimental results demonstrate our proposed sequence modeling architecture design principle and suggest that our work can serve as a valuable reference for future architecture improvements and novel designs.

Multilingual reasoning remains a significant challenge for large language models (LLMs), with performance disproportionately favoring high-resource languages. Drawing inspiration from cognitive neuroscience, which suggests that human reasoning functions largely independently of language processing, we hypothesize that LLMs similarly encode reasoning and language as separable components that can be disentangled to enhance multilingual reasoning. To evaluate this, we perform a causal intervention by ablating language-specific representations at inference time. Experiments on 10 open-weight LLMs spanning 11 typologically diverse languages show that this language-specific ablation consistently boosts multilingual reasoning performance. Layer-wise analyses further confirm that language and reasoning representations can be effectively disentangled throughout the model, yielding improved multilingual reasoning capabilities, while preserving top-layer language features remains essential for maintaining linguistic fidelity. Compared to post-training methods such as supervised fine-tuning or reinforcement learning, our training-free language-reasoning disentanglement achieves comparable or superior results with minimal computational overhead. These findings shed light on the internal mechanisms underlying multilingual reasoning in LLMs and suggest a lightweight and interpretable strategy for improving cross-lingual generalization.

Personalized alignment is essential for enabling large language models (LLMs) to engage effectively in user-centric dialogue. While recent prompt-based and offline optimization methods offer preliminary solutions, they fall short in cold-start scenarios and long-term personalization due to their inherently static and shallow designs. In this work, we introduce the Reinforcement Learning for Personalized Alignment (RLPA) framework, in which an LLM interacts with a simulated user model to iteratively infer and refine user profiles through dialogue. The training process is guided by a dual-level reward structure: the Profile Reward encourages accurate construction of user representations, while the Response Reward incentivizes generation of responses consistent with the inferred profile. We instantiate RLPA by fine-tuning Qwen-2.5-3B-Instruct, resulting in Qwen-RLPA, which achieves state-of-the-art performance in personalized dialogue. Empirical evaluations demonstrate that Qwen-RLPA consistently outperforms prompting and offline fine-tuning baselines, and even surpasses advanced commercial models such as Claude-3.5 and GPT-4o. Further analysis highlights Qwen-RLPA's robustness in reconciling conflicting user preferences, sustaining long-term personalization and delivering more efficient inference compared to recent reasoning-focused LLMs. These results emphasize the potential of dynamic profile inference as a more effective paradigm for building personalized dialogue systems.

A primary impediment to scaling reinforcement learning (RL) for large language model (LLM) training is the substantial computational cost, predominantly arising from the necessity of multi-sampling for policy optimization and evaluation. This underscores the critical yet challenging nature of efficient training data selection. Drawing inspiration from the Zone of Proximal Development (ZPD) theory, which posits that learners acquire knowledge more effectively from tasks of intermediate difficulty, we hypothesize that LLMs exhibit optimal learning from data they have not yet mastered but demonstrate the potential to comprehend. Conventional methodologies for assessing data difficulty or informativeness typically rely on computationally intensive multi-sampling or iterative procedures. To address this limitation, we introduce UFO-RL (**U**ncertainty-**F**ocused **O**ptimization for **R**einforcement **L**earning), a novel framework that employs a computationally efficient single-pass uncertainty estimation technique to identify informative training instances. This method, requiring only a single forward pass and obviating the need for iterative next-token computation, achieves a significant acceleration (up to 185$\times$) in data evaluation compared to multi-sampling approaches. UFO-RL leverages this efficient metric to select data within the model's estimated ZPD for training. Extensive experimentation across diverse LLMs and mathematical benchmarks demonstrates that training with a mere 10\% of the data, carefully selected by UFO-RL, yields performance comparable to or even surpassing that of full-data training. Furthermore, this targeted data selection results in up to a 16$\times$ reduction in overall training time, concurrently enhancing training stability and improving generalization capabilities. Thus, UFO-RL presents a practical and highly efficient strategy for scaling RL fine-tuning of LLMs by focusing learning efforts on the most informative and valuable data, thereby mitigating the computational bottlenecks associated with traditional RL training.

Recent advancements in dialogue policy planning have focused on optimizing system agent policies to achieve predefined goals, emphasizing strategy design, trajectory acquisition, and training efficiency.However, these approaches often overlook the critical role of user characteristics, which are essential in real-world scenarios like conversational search and recommendation, where interactions must adapt to individual user traits such as personality, preferences, and goals.To address this gap, we conduct a comprehensive study using task-specific user personas to evaluate dialogue policy planning under diverse user behaviors.Our analysis, based on these user profiles, reveals significant shortcomings in existing approaches, underscoring the necessity for user-tailored dialogue policies.Building on these insights, we propose the User-Tailored Dialogue Policy Planning (UDP) framework, which integrates an Intrinsic User World Model to capture user traits and feedback.UDP operates in three stages: (1) User Persona Portraying, employing a diffusion model to dynamically infer user profiles; (2) User Feedback Anticipating, using a Brownian Bridge-inspired mechanism to predict user reactions; and (3) User-Tailored Policy Planning, synthesizing these elements to optimize response strategies.To enhance robustness, we introduce an active learning approach that prioritizes challenging user personas during training.Extensive experiments across benchmarks, including both collaborative and non-collaborative settings, demonstrate UDP's effectiveness in learning user-specific dialogue strategies.Results confirm the framework's utility, highlighting its robustness, adaptability, and potential to advance user-centric dialogue systems.

Multimodal Emotion Recognition in Conversations (MERC) aims to identify the emotion expressed in each utterance within conversational videos. Current efforts are directed toward modeling speaker-sensitive context dependencies and multimodal fusion. However, they still struggle to handle utterances from unseen speakers, hampering the model's generalizability. To tackle this challenge, we propose a Speaker Generalization Framework for MERC. Specifically, we build a prototype graph to learn Speaker-based Utterance Representations (SUR), leveraging prototypes as the bridge between seen and unseen speakers. Speaker-aware Contrastive Learning (CL) is then applied to refine SUR, pulling utterances (or prototypes) from the same speaker together while pushing those from different speakers apart. Further, we introduce a prototypical graph CL to generalize SUR to unseen speakers, ensuring that the same speakers exhibit similar graph structures, while dissimilar ones differ. To further enhance model generalization, we introduce Uncertainty-based Generalization for Speakers, randomly sampling SUR statistics from the estimated Gaussian distribution and probabilistically replacing the original SUR. Experimental findings on two datasets highlight that our framework substantially improves the generalization of various MERC models, surpassing state-of-the-art methods.

Multimodal sentiment analysis (MSA) aims to leverage the complementary information from multiple modalities for affective understanding of user-generated videos. Existing methods mainly focused on designing sophisticated feature fusion strategies to integrate the separately extracted multimodal representations, ignoring the interference of the information irrelevant to sentiment. In this paper, we propose to disentangle the unimodal representations into sentiment-specific and sentiment-independent features, the former of which are fused for the MSA task. Specifically, we design a novel Sentiment-aware Disentangled Representation Shifting framework, termed SDRS, with two components.Interactive sentiment-aware representation disentanglementaims to extract sentiment-specific feature representations for each nonverbal modality by considering the contextual influence of other modalities with the newly developed cross-attention autoencoder.Attentive cross-modal representation shiftingtries to shift the textual representation in a latent token space using the nonverbal sentiment-specific representations after projection. The shifted representation is finally employed to fine-tune a pre-trained language model for multimodal sentiment analysis. Extensive experiments are conducted on three public benchmark datasets, i.e., CMU-MOSI, CMU-MOSEI, and CH-SIMS. The results demonstrate that the proposed SDRS framework not only obtains state-of-the-art results based solely on multimodal labels but also outperforms the methods that additionally require the labels of each modality.

This study investigates a novel commonsense inference task that comprises reasoning about the subsequent events chains and probable consumption intents, given an event described in brief free-form text. For instance, given the event “fall in love”, the system anticipates ensuing events such as “get married”→“be pregnant”→“have a baby”. Concurrently, the system predicts the likely consumption intents of the event participants, such as purchasing “jewelry”, “maternity clothing”, and “baby food”. The event chains generated in this process provide explicit, meaningful explanations, aiding the understanding of inferred consumption intents. To facilitate this study, we construct a new crowdsourced corpus: x-ECON. This corpus comprises 10,144 event chains and 150 event-related product categories, offering a wide range of everyday consumer events and situations. We introduce a baseline reasoning framework that not only infers consumption intent but also generates an event chain to explain its inference. Our experimental results suggest that involving the event chain reasoning in the event-consumption reasoning system can help improve the neural networks' reason about the probable consumption intents of the event participants. Additionally, our method demonstrates the applicability of our approach in improving the performance of recommendation systems, by highlighting the role of explainable commonsense inference on the consumption intent. We also evaluated the performance of ChatGPT on our x-ECON dataset, showing that explainable event-consumption commonsense reasoning is a challenging task for large language models.

Large Language Models (LLMs) have demonstrated remarkable success in diverse natural language processing (NLP) tasks in general domains. However, LLMs sometimes generate responses with the hallucination about medical facts due to limited domain knowledge. Such shortcomings pose potential risks in the utilization of LLMs within medical contexts. To address this challenge, we propose knowledge-tuning, which leverages structured medical knowledge bases for the LLMs to grasp domain knowledge efficiently and facilitate trustworthy response generation. We also release cMedKnowQA, a Chinese medical knowledge question-answering dataset constructed from medical knowledge bases to assess the medical knowledge proficiency of LLMs. Experimental results show that the LLMs which are knowledge-tuned with cMedKnowQA can exhibit higher levels of accuracy in response generation compared with vanilla instruction-tuning and offer a new trustworthy way for the domain adaptation of LLMs. We release our code and data at https://github.com/SCIR-HI/Huatuo-Llama-Med-Chinese .

Many question-answering problems can be approached as textual entailment tasks, where the hypotheses are formed by the question and candidate answers, and the premises are derived from an external knowledge base. However, current neural methods often lack transparency in their decision-making processes. Moreover, first-order logic methods, while systematic, struggle to integrate unstructured external knowledge. To address these limitations, we propose a neuro-symbolic reasoning framework calledFinal, which combinesFIrst-order logic withNAturalLogic for question answering. Our framework utilizesfirst-order logicto systematically decompose hypotheses andnatural logicto construct reasoning paths from premises to hypotheses, employing bidirectional reasoning to establish links along the reasoning path. This approach not only enhances interpretability but also effectively integrates unstructured knowledge. Our experiments on three benchmark datasets, namely QASC, WorldTree, and WikiHop, demonstrate thatFinaloutperforms existing methods in commonsense reasoning and reading comprehension tasks, achieving state-of-the-art results. Additionally, our framework also provides transparent reasoning paths that elucidate the rationale behind the correct decisions.

Deep learning (DL) is increasingly viewed as a foundational methodology for advancing Artificial Intelligence (AI). However, its interpretability remains limited, and it often underperforms in certain fields due to its lack of human-like characteristics. Consequently, leveraging insights from Brain and Cognitive Science (BCS) to understand and advance DL has become a focal point for researchers in the DL community. However, BCS is a diverse discipline where existing studies often concentrate on cognitive theories within their respective domains. These theories are typically grounded in certain assumptions, complicating comparisons between different approaches. Therefore, this review is intended to provide a comprehensive landscape of more than 300 papers on the intersection of DL and BCS grounded in DL community. Unlike previous reviews that based on sub-disciplines of Cognitive Science, this article aims to establish a unified framework encompassing all aspects of DL inspired by BCS, offering insights into the symbiotic relationship between DL and BCS. Additionally, we present a forward-looking perspective on future research directions, with the intention of inspiring further advancements in AI research.

The emergence of large language models (LLMs) has marked a significant breakthrough in natural language processing (NLP), fueling a paradigm shift in information acquisition. Nevertheless, LLMs are prone to hallucination, generating plausible yet nonfactual content. This phenomenon raises significant concerns over the reliability of LLMs in real-world information retrieval (IR) systems and has attracted intensive research to detect and mitigate such hallucinations. Given the open-ended general-purpose attributes inherent to LLMs, LLM hallucinations present distinct challenges that diverge from prior task-specific models. This divergence highlights the urgency for a nuanced understanding and comprehensive overview of recent advances in LLM hallucinations. In this survey, we begin with an innovative taxonomy of hallucination in the era of LLM and then delve into the factors contributing to hallucinations. Subsequently, we present a thorough overview of hallucination detection methods and benchmarks. Our discussion then transfers to representative methodologies for mitigating LLM hallucinations. Additionally, we delve into the current limitations faced by retrieval-augmented LLMs in combating hallucinations, offering insights for developing more robust IR systems. Finally, we highlight the promising research directions on LLM hallucinations, including hallucination in large vision-language models and understanding of knowledge boundaries in LLM hallucinations.

Molecule discovery serves as a cornerstone in numerous scientific domains, fueling the development of new materials and innovative drug designs. Recent developments of in-silico molecule discovery have highlighted the promising results of cross-modal techniques, which bridge molecular structures with their descriptive annotations. However, these cross-modal methods frequently encounter the issue of data scarcity, hampering their performance and application. In this paper, we address the low-resource challenge by utilizing artificially-real data generated by Large Language Models (LLMs). We first introduce a retrieval-based prompting strategy to construct high-quality pseudo data, then explore the optimal method to effectively leverage this pseudo data. Experiments show that using pseudo data for domain adaptation outperforms all existing methods, while also requiring a smaller model scale, reduced data size and lower training cost, highlighting its efficiency. Furthermore, our method shows a sustained improvement as the volume of pseudo data increases, revealing the great potential of pseudo data in advancing low-resource cross-modal molecule discovery.

Deep learning is now widely used in drug discovery, providing significant acceleration and cost reduction. As the most fundamental building block, molecular representation is essential for predicting molecular properties to enable various downstream applications. Most existing methods attempt to incorporate more information to learn better representations. However, not all features are equally important for a specific task. Ignoring this would potentially compromise the training efficiency and predictive accuracy. To address this issue, we propose a novel approach, which treats language models as an agent and molecular pretraining models as a knowledge base. The agent accentuates task-relevant features in the molecular representation by understanding the natural language description of the task, just as a tailor customizes clothes for clients. Thus, we call this approach MolTailor. Evaluations demonstrate MolTailor's superior performance over baselines, validating the efficacy of enhancing relevance for molecular representation learning. This illustrates the potential of language model guided optimization to better exploit and unleash the capabilities of existing powerful molecular representation methods. Our code and appendix are available at https://github.com/SCIR-HI/MolTailor.

Prompt-based classification adapts tasks to a cloze question format utilizing the [MASK] token and the filled tokens are then mapped to labels through pre-defined verbalizers. Recent studies have explored the use of verbalizer embeddings to reduce labor in this process. However, all existing studies require a tuning process for either the pre-trained models or additional trainable embeddings. Meanwhile, the distance between high-dimensional verbalizer embeddings should not be measured by Euclidean distance due to the potential for non-linear manifolds in the representation space. In this study, we propose a tuning-free manifold-based space re-embedding method called Locally Linear Embedding with Intra-class Neighborhood Constraint (LLE-INC) for verbalizer embeddings, which preserves local properties within the same class as guidance for classification. Experimental results indicate that even without tuning any parameters, our LLE-INC is on par with automated verbalizers with parameter tuning. And with the parameter updating, our approach further enhances prompt-based tuning by up to 3.2%. Furthermore, experiments with the LLaMA-7B&13B indicate that LLE-INC is an efficient tuning-free classification approach for the hyper-scale language models.

Kun Zhu, Xiaocheng Feng, Xiyuan Du, Yuxuan Gu, Weijiang Yu, Haotian Wang, Qianglong Chen, Zheng Chu, Jingchang Chen, Bing Qin. Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2024.

Zheng Chu, Jingchang Chen, Qianglong Chen, Haotian Wang, Kun Zhu, Xiyuan Du, Weijiang Yu, Ming Liu, Bing Qin. Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2024.

Zheng Chu, Jingchang Chen, Qianglong Chen, Weijiang Yu, Haotian Wang, Ming Liu, Bing Qin. Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2024.

Zheng Chu, Jingchang Chen, Qianglong Chen, Weijiang Yu, Tao He, Haotian Wang, Weihua Peng, Ming Liu, Bing Qin, Ting Liu. Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2024.

Jinglong Gao, Xiao Ding, Yiming Cui, Jianbai Zhao, Hepeng Wang, Ting Liu, Bing Qin. Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2024.

In proactive dialogue, the challenge lies not just in generating responses but in steering conversations toward predetermined goals, a task where Large Language Models (LLMs) typically struggle due to their reactive nature.Traditional approaches to enhance dialogue planning in LLMs, ranging from elaborate prompt engineering to the integration of policy networks, either face efficiency issues or deliver suboptimal performance.Inspired by the dualprocess theory in psychology, which identifies two distinct modes of thinking-intuitive (fast) and analytical (slow), we propose the Dual-Process Dialogue Planning (DPDP) framework.DPDP embodies this theory through two complementary planning systems: an instinctive policy model for familiar contexts and a deliberative Monte Carlo Tree Search (MCTS) mechanism for complex, novel scenarios.This dual strategy is further coupled with a novel two-stage training regimen: offline Reinforcement Learning for robust initial policy model formation followed by MCTS-enhanced on-thefly learning, which ensures a dynamic balance between efficiency and strategic depth.Our empirical evaluations across diverse dialogue tasks affirm DPDP's superiority in achieving both high-quality dialogues and operational efficiency, outpacing existing methods. 1

Zhouhao Sun, Li Du, Xiao Ding, Yixuan Ma, Yang Zhao, Kaitao Qiu, Ting Liu, Bing Qin. Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2024.

Weixiang Zhao, Shilong Wang, Yulin Hu, Yanyan Zhao, Bing Qin, Xuanyu Zhang, Qing Yang, Dongliang Xu, Wanxiang Che. Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2024.

Weihong Zhong, Xiaocheng Feng, Liang Zhao, Qiming Li, Lei Huang, Yuxuan Gu, Weitao Ma, Yuan Xu, Bing Qin. Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2024.

With the development of Large Language Models (LLMs), many Chinese medical benchmarks have emerged. These benchmarks have primarily used multiple-choice questions and open-ended questions as test items. However, our experimental results indicate that using multiple-choice questions to test the capabilities of LLMs is not very reasonable. Additionally, relatively simple open-ended questions do not effectively assess LLMs’ actual grasp of medical knowledge. Therefore, we propose the Chinese Medical Complex Open-Question Answering Benchmark (CMCOQA), designed to more accurately and efficiently evaluate the true medical proficiency of LLMs by constructing complex open-ended questions within medical scenarios. Our proposed benchmark involves three evaluation dimensions: Completeness, Depth, and Professionalism. Starting with 100 manually generated complex questions as seeds, we expand the set to 1,200 using the Self-Instruct method with GPT-4o. We then have GPT-4o self-check the questions, followed by a manual screening process to ensure a broad coverage and a certain level of depth. We have both humans and GPT-4o score from these three dimensions, while also employing automated metrics. We also calculate correlations between these metrics and human scores to validate the results. Through this work, CMCOQA can further promote the development of Chinese medical LLMs in terms of medical professionalism.

Large Language Models (LLMs) are gaining widespread attention for their potential across various fields, particularly within the medical domain. Recent efforts have aimed to privatize general-domain LLMs into specialized medical-domain LLMs by feeding high-quality, medical-domain training data. However, an overlooked aspect in these privatization efforts is the Dual Logic Ability of LLMs. This ability enables LLMs to comprehend pairs of logically opposed questions, ensuring stance consistency in their responses. In our study, we investigate two primary questions: Q1. How does privatization affect the dual logic ability of LLMs? Q2. How can we maintain the robustness of LLMs’ dual logic ability after privatization? To explore these questions, we first constructed a medical-domain dual logic ability evaluation dataset comprising logically opposed question pairs, created manually by NLP experts. By examining the stance consistency in responses to logically opposed question pairs, our analysis demonstrates a significant decline in the dual logic ability of LLMs after privatization. Furthermore, we construct privatization data to investigate the effects of the pre-training and instruction fine-tuning stages on the dual logic ability of LLMs. Interestingly, our findings reveal that the instruction fine-tuning stage often inadvertently compromises the LLMs’ dual logic ability although it is not the trainers’ intention. To counteract this, we incorporated general-domain dual logic data derived from basic science during the instruction fine-tuning stage, which are automatically constructed by our designed pipeline. Experiment results show that privatized LLMs can generalize dual logic ability from general-domain dual logic data, leading to their enhanced performance in the medical domain. Our study underscores the importance of prioritizing LLMs’ dual logic ability during the privatization process and establishes a benchmark for future research. Our data and code are available at GitHub1.

Mental health issues pose significant challenges worldwide, leading researchers to utilize computing techniques to analyze social media posts for early detection. Currently, Large Language Models (LLMs) represent advanced AI systems with exceptional language processing capabilities. However, their effectiveness depends on the use of appropriate prompts. The design of suitable prompts specifically for analyzing mental health issues remains underexplored. Drawing inspiration from psychiatrists who use psychiatric scales to assess mental health, we have developed scale-based chains of thought (Scale-CoT) prompts. These not only enhance the LLMs’ ability to detect mental health issues but also increase clarity and explainability by integrating professional scales. Furthermore, we demonstrate that the knowledge generated by LLMs can be effectively transferred to smaller models, resulting in a better-performing and more lightweight system. Our approach, which leverages psychiatric scales, outperforms existing methods. Our experiments indicate that combining traditional psychiatry with computational techniques is worthwhile and deserving of further research and exploration.

Multilingual neural machine translation handles the translation of multiple languages with one unified model. However, this joint-training paradigm incurs the notorious issue of parameter interference, where the model compromises with the language diversity to find a common solution. Recent research has explored avoiding this problem by selecting certain parameters for each language direction from the original model to form language-specific sub-networks. However, determining how many parameters to choose and which parameters to select is still a serious challenge. In this work, we propose an approach called CaPA (Consistency-based Parameter Allocation), which dynamically allocates parameters of appropriate scale to each language direction based on the consistency between the gradient of the individual language and the average gradient. Specifically, CaPA allocates more parameters to languages with higher gradient consistency as these languages tend to have a more positive impact on other languages. Furthermore, considering the varying levels of interference across different parts of the model, we propose an adaptive parameter allocation based on module-level gradient consistency. Experimental results show the correlation between gradient consistency and parameter interference, as well as the effectiveness of our proposed method.

Large language models (LLMs) have achieved significant performance in various natural language reasoning tasks. However, they still struggle with performing first-order logic reasoning over formal logical theories expressed in natural language. This is because the previous LLMs-based reasoning systems have the theoretical incompleteness issue. As a result, it can only address a limited set of simple reasoning problems, which significantly decreases their generalization ability. To address this issue, we propose a novel framework, named Generalizable and Faithful Reasoner (GFaiR), which introduces the paradigm of resolution refutation. Resolution refutation has the capability to solve all first-order logic reasoning problems by extending reasoning rules and employing the principle of proof by contradiction, so our system’s completeness can be improved by introducing resolution refutation. Experimental results demonstrate that our system outperforms previous works by achieving state-of-the-art performances in complex scenarios while maintaining performances in simple scenarios. Besides, we observe that GFaiR is faithful to its reasoning process.

Despite achieving remarkable performance on various vision-language tasks, Transformer-based Vision-Language Models (VLMs) suffer from redundancy in inputs and parameters, significantly hampering their efficiency in real-world applications. Moreover, the degree of redundancy in token representations and model parameters, such as attention heads, varies significantly for different inputs. In light of the challenges, we propose SmartTrim, an adaptive acceleration framework for VLMs, which adjusts the computational overhead per instance. Specifically, we integrate lightweight modules into the original backbone to identify and prune redundant token representations and attention heads within each layer. Furthermore, we devise a self-distillation strategy to enhance the consistency between the predictions of the pruned model and its fully-capacity counterpart. Experimental results across various vision-language tasks consistently demonstrate that SmartTrim accelerates the original model by 2-3 times with minimal performance degradation, highlighting the effectiveness and efficiency compared to previous approaches. Code will be available at https://github.com/kugwzk/SmartTrim.

Depression impacts millions worldwide, with increasing efforts to use social media data for early detection and intervention. Traditional Risk Detection (TRD) uses a user’s complete posting history for predictions, while Early Risk Detection (ERD) seeks early detection in a user’s posting history, emphasizing the importance of prediction earliness. However, ERD remains relatively underexplored due to challenges in balancing accuracy and earliness, especially with evolving partial data. To address this, we introduce the Early Sensing Depression Model (ESDM), which comprises two modules classification with partial information module (CPI) and decision for classification moment module (DMC), alongside an early detection loss function. Experiments show ESDM outperforms benchmarks in both earliness and accuracy.

Fact-checking is the task of verifying the factuality of a given claim by examining the available evidence. High-quality evidence plays a vital role in enhancing fact-checking systems and facilitating the generation of explanations that are understandable to humans. However, the provision of both sufficient and relevant evidence for explainable fact-checking systems poses a challenge. To tackle this challenge, we propose a method based on a Large Language Model to automatically retrieve and summarize evidence from the Web. Furthermore, we construct RU22Fact, a novel multilingual explainable fact-checking dataset on the Russia-Ukraine conflict in 2022 of 16K samples, each containing real-world claims, optimized evidence, and referenced explanation. To establish a baseline for our dataset, we also develop an end-to-end explainable fact-checking system to verify claims and generate explanations. Experimental results demonstrate the prospect of optimized evidence in increasing fact-checking performance and also indicate the possibility of further progress in the end-to-end claim verification and explanation generation tasks.

Multi-modal Named Entity Recognition, a fundamental task for multi-modal knowledge graph construction, requires integrating multi-modal information to extract named entities from text. Previous research has explored the integration of multi-modal representations at different granularities. However, they struggle to integrate all these multi-modal representations to provide rich contextual information to improve multi-modal named entity recognition. In this paper, we propose DPE-MNER, which is an iterative reasoning framework that dynamically incorporates all the diverse multi-modal representations following the strategy of “decompose, prioritize, and eliminate”. Within the framework, the fusion of diverse multi-modal representations is decomposed into hierarchically connected fusion layers that are easier to handle. The incorporation of multi-modal information prioritizes transitioning from “easy-to-hard” and “coarse-to-fine”. The explicit modeling of cross-modal relevance eliminate the irrelevances that will mislead the MNER prediction. Extensive experiments on two public datasets have demonstrated the effectiveness of our approach.

Lei Huang, Xiaocheng Feng, Weitao Ma, Liang Zhao, Yuchun Fan, Weihong Zhong, Dongliang Xu, Qing Yang, Hongtao Liu, Bing Qin. Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing. 2024.

Despite the remarkable progress made by large language models in mathematical reasoning, interactive theorem proving in formal logic still remains a prominent challenge.Previous methods resort to neural models for proofstep generation and search.However, they suffer from exploring possible proofsteps empirically in a large search space.Besides, they directly use a less rigorous informal proof for proofstep generation, neglecting the incomplete reasoning within.In this paper, we propose BC-Prover, a backward chaining framework guided by pseudo steps.Specifically, BC-Prover prioritizes pseudo steps to proofstep generation.The pseudo steps boost the proof construction in two aspects: (1) Backward Chaining that decomposes the proof into sub-goals for goaloriented exploration.(2) Step Planning that makes a fine-grained planning to bridge the gap between informal and formal proofs.Experiments on the miniF2F benchmark show significant performance gains by our framework over the state-of-the-art approaches.Our framework is also compatible with existing provers and further improves their performance with the backward chaining technique.

Large Language models (LLMs) have exhibited remarkable abilities in understanding complex texts, offering a promising path towards human-like translation performance.However, this study reveals the misalignment between the translation-specific understanding and the general understanding inside LLMs.This understanding misalignment leads to LLMs mistakenly or literally translating some complicated concepts that they accurately comprehend in the general scenarios (e.g., QA).To align the translation-specific understanding to the general one, we propose a novel translation process, DUAT (Difficult words Understanding Aligned Translation), explicitly incorporating the general understanding on the complicated content incurring inconsistent understanding to guide the translation.Specifically, DUAT performs cross-lingual interpretation for the difficult-to-translate words and enhances the translation with the generated interpretations.Furthermore, we reframe the external tools to improve DUAT in detecting difficult words and generating helpful interpretations.We conduct experiments on the self-constructed benchmark Challenge-WMT 1 , consisting of samples that are prone to mistranslation.Human evaluation results on highresource and low-resource language pairs indicate that DUAT significantly facilitates the understanding alignment, which improves the translation quality (up to +3.85 COMET) and reduces the literality of the translation by -25% ∼ -51%.

Scaling the rotary position embedding (RoPE) has become a common method for extending the context window of RoPE-based large language models (LLMs).However, existing scaling methods often rely on empirical approaches and lack a profound understanding of the internal distribution within RoPE, resulting in suboptimal performance in extending the context window length.In this paper, we propose to optimize the context window extending task from the view of rotary angle distribution.Specifically, we first estimate the distribution of the rotary angles within the model and analyze the extent to which length extension perturbs this distribution.Then, we present a novel extension strategy that minimizes the disturbance between rotary angle distributions to maintain consistency with the pre-training phase, enhancing the model's capability to generalize to longer sequences.Experimental results compared to the strong baseline methods demonstrate that our approach reduces by up to 72% of the distributional disturbance when extending LLaMA2's context window to 8k, and reduces by up to 32% when extending to 16k.On the LongBench-E benchmark, our method achieves an average improvement of up to 4.33% over existing state-of-the-art methods.Furthermore, our method maintains the model's performance on the Hugging Face Open LLM benchmark after context window extension, with only an average performance fluctuation ranging from -0.12 to +0.22.Our code is available at https: //github.com/1180301012/DPRoPE.

Yangfan Ye, Xiachong Feng, Xiaocheng Feng, Weitao Ma, Libo Qin, Dongliang Xu, Qing Yang, Hongtao Liu, Bing Qin. Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing. 2024.

Traditional multitask learning methods typically can only leverage shared knowledge within specific tasks or languages, resulting in a loss of either cross-language or cross-task knowledge. This paper proposes a general multilingual multitask model, named SkillNet-X, which enables a single model to tackle many different tasks from different languages. To this end, we define several language-specific skills and task-specific skills, each of which corresponds to a skill module. SkillNet-X sparsely activates parts of the skill modules which are relevant to eitherthe target task or the target language. Acting as knowledge transit hubs, skill modules are capable of absorbing task-related knowledge and language-related knowledge consecutively. We evaluate SkillNet-X on eleven natural language understanding datasets in four languages. Results show that SkillNet-X performs better than task-specific and two multitask learning baselines.To investigate the generalization of our model, we conduct experiments on two new tasks and find that SkillNet-X significantly outperforms baselines.

Large language models augmented with task-relevant documents have demonstrated impressive performance on knowledge-intensive tasks. However, regarding how to obtain effective documents, the existing methods are mainly divided into two categories. One is to retrieve from an external knowledge base, and the other is to utilize large language models to generate documents. We propose an iterative retrieval-generation collaborative framework. It is not only able to leverage both parametric and non-parametric knowledge, but also helps to find the correct reasoning path through retrieval-generation interactions, which is very important for tasks that require multi-step reasoning. We conduct experiments on four question answering datasets, including single-hop QA and multi-hop QA tasks. Empirical results show that our method significantly improves the reasoning ability of large language models and outperforms previous baselines.

Multi-modal relation extraction (MRE) requires the integration of multi-modal information to identify relationships between entities. Although fine-grained correlations between visual objects and textual words have the potential to improve cross-modal interaction, they are typically modeled implicitly and hindered by the modality gap. This paper introduces a novel method called relational Graph-Bridged cross-modal InTeraction (GBIT). GBIT aims to model fine-grained cross-modal correlations into the interaction process explicitly. This is achieved by constructing a fine-grained cross-modal relational graph, which acts as a bridge for effective cross-modal interaction in multiple layers. Within GBIT, a gated interaction strategy and an adaptive integration module are proposed for irrelevance-filtered information exchange and final information collation. Through extensive experiments on the benchmark MRE, we demonstrate the superiority of our proposed method for MRE.

The objective of suicide risk detection on social media is to identify individuals who may attempt suicide and determine their suicide risk level based on their online behavior. Although data-driven learning models have been used to predict suicide risk levels, these models often lack theoretical support and explanation from psychiatric research. To address this issue, we propose the incorporation of professional psychiatric scales into research to provide theoretical support and explanations for our model. Our proposed Scale-based Neural Network (SNN) architecture aims to extract content associated with scales from the posting history of social media users to predict their suicide risk level. Additionally, our approach provides scale-based explanations for the model's predictions. Experimental results demonstrate that our proposed method outperforms several strong baseline methods and highlights the potential of combining psychiatric scales and computational techniques to improve suicide risk detection.

Large language models (LLMs) have developed impressive performance and strong explainability across various reasoning scenarios, marking a significant stride towards mimicking human-like intelligence. Despite this, when tasked with several simple questions supported by a generic fact, LLMs often struggle to abstract and apply the generic fact to provide consistent and precise answers, revealing a deficiency in abstract reasoning abilities. This has sparked a vigorous debate about whether LLMs are genuinely reasoning or merely memorizing. In light of this, we design a preliminary study to quantify and delve into the abstract reasoning abilities of existing LLMs. Our findings reveal a substantial discrepancy between their general reasoning and abstract reasoning performances. To relieve this problem, we tailor an abstract reasoning dataset (AbsR) together with a meaningful learning paradigm to teach LLMs how to leverage generic facts for reasoning purposes. The results show that our approach not only boosts the general reasoning performance of LLMs but also makes considerable strides towards their capacity for abstract reasoning, moving beyond simple memorization or imitation to a more nuanced understanding and application of generic facts. The code is available at https://github.com/Waste-Wood/MeanLearn.

Despite recent progress made by large language models in code generation, they still struggle with programs that meet complex requirements. Recent work utilizes plan-and-solve decomposition to decrease the complexity and leverage self-tests to refine the generated program. Yet, planning deep-inside requirements in advance can be challenging, and the tests need to be accurate to accomplish self-improvement. To this end, we propose FunCoder, a code generation framework incorporating the divide-and-conquer strategy with functional consensus. Specifically, FunCoder recursively branches off sub-functions as smaller goals during code generation, represented by a tree hierarchy. These sub-functions are then composited to attain more complex objectives. Additionally, we designate functions via a consensus formed by identifying similarities in program behavior, mitigating error propagation. FunCoder outperforms state-of-the-art methods by +9.8% on average in HumanEval, MBPP, xCodeEval and MATH with GPT-3.5 and GPT-4. Moreover, our method demonstrates superiority on smaller models: With FunCoder, StableCode-3b surpasses GPT-3.5 by +18.6% and achieves 97.7% of GPT-4's performance on HumanEval. Further analysis reveals that our proposed dynamic function decomposition is capable of handling complex requirements, and the functional consensus prevails over self-testing in correctness evaluation.

Large Language Models (LLMs) are increasingly deployed in various applications. As their usage grows, concerns regarding their safety are rising, especially in maintaining harmless responses when faced with malicious instructions. Many defense strategies have been developed to enhance the safety of LLMs. However, our research finds that existing defense strategies lead LLMs to predominantly adopt a rejection-oriented stance, thereby diminishing the usability of their responses to benign instructions. To solve this problem, we introduce the MoGU framework, designed to enhance LLMs' safety while preserving their usability. Our MoGU framework transforms the base LLM into two variants: the usable LLM and the safe LLM, and further employs dynamic routing to balance their contribution. When encountering malicious instructions, the router will assign a higher weight to the safe LLM to ensure that responses are harmless. Conversely, for benign instructions, the router prioritizes the usable LLM, facilitating usable and helpful responses. On various open-sourced LLMs, we compare multiple defense strategies to verify the superiority of our MoGU framework. Besides, our analysis provides key insights into the effectiveness of MoGU and verifies that our designed routing mechanism can effectively balance the contribution of each variant by assigning weights. Our work released the safer Llama2, Vicuna, Falcon, Dolphin, and Baichuan2.

We present an novel framework for efficiently and effectively extending the powerful continuous diffusion processes to discrete modeling. Previous approaches have suffered from the discrepancy between discrete data and continuous modeling. Our study reveals that the absence of guidance from discrete boundaries in learning probability contours is one of the main reasons. To address this issue, we propose a two-step forward process that first estimates the boundary as a prior distribution and then rescales the forward trajectory to construct a boundary conditional diffusion model. The reverse process is proportionally adjusted to guarantee that the learned contours yield more precise discrete data. Experimental results indicate that our approach achieves strong performance in both language modeling and discrete image generation tasks. In language modeling, our approach surpasses previous state-of-the-art continuous diffusion language models in three translation tasks and a summarization task, while also demonstrating competitive performance compared to auto-regressive transformers. Moreover, our method achieves comparable results to continuous diffusion models when using discrete ordinal pixels and establishes a new state-of-the-art for categorical image generation on the Cifar-10 dataset.

Large language models (LLMs) exhibit complementary strengths in various tasks, motivating the research of LLM ensembling. However, existing work focuses on training an extra reward model or fusion model to select or combine all candidate answers, posing a great challenge to the generalization on unseen data distributions. Besides, prior methods use textual responses as communication media, ignoring the valuable information in the internal representations. In this work, we propose a training-free ensemble framework \textsc{DeePEn}, fusing the informative probability distributions yielded by different LLMs at each decoding step. Unfortunately, the vocabulary discrepancy between heterogeneous LLMs directly makes averaging the distributions unfeasible due to the token misalignment. To address this challenge, \textsc{DeePEn} maps the probability distribution of each model from its own probability space to a universal \textit{relative space} based on the relative representation theory, and performs aggregation. Next, we devise a search-based inverse transformation to transform the aggregated result back to the probability space of one of the ensembling LLMs (main model), in order to determine the next token. We conduct extensive experiments on ensembles of different number of LLMs, ensembles of LLMs with different architectures, and ensembles between the LLM and the specialist model. Experimental results show that (i) \textsc{DeePEn} achieves consistent improvements across six benchmarks covering subject examination, reasoning, and knowledge, (ii) a well-performing specialist model can benefit from a less effective LLM through distribution fusion, and (iii) \textsc{DeePEn} has complementary strengths with other ensemble methods such as voting.

Pre-trained language models have been proven to possess strong base capabilities, which not only excel in in-distribution language modeling but also show powerful abilities in out-of-distribution language modeling, transfer learning and few-shot learning. Unlike existing work focusing on the influence of scale on base capabilities, our work examines the influence of architecture on those. Specifically, our concern is: How does architecture influence the base capabilities of pre-trained language models? In this work, we attempt to explain and reverse the decline in base capabilities caused by the architecture of FFN-Wider Transformers, seeking to provide some insights. Through analysis, we found the contribution ratio of Multi-Head Attention (a combination function) to pre-trained language modeling is a key factor affecting base capabilities. FFN-Wider Transformers reduce the contribution ratio of this combination function, leading to a decline in base capabilities. We confirmed this by experiments and proposed Combination Enhanced Architecture (CEA) to address the decline in base capabilities of such models. Significantly, we extended our explanation and CEA to Mixture of Experts (MoE) Transformers. We successfully achieved significant improvements in base capabilities on a 14B parameter MoE model, demonstrating the practical application value of our work. This also indicates that our analysis has a certain guiding significance for architecture analysis, architecture improvement and architecture design.

Given a text-image pair, Multimodal Named Entity Recognition (MNER) is the task of identifying and categorizing entities in the text. Most existing work performs named entity labeling directly using final token representations derived by fusing image and text representations. Although they achieve promising results, these work may fail to effectively exploit text and image modalities. This is because they neglect the difference in the role of the two modalities: text modality can detect the boundary of an entity, while image modality is introduced to disambiguate the category of the entity. Based on these findings, in this paper, we construct two auxiliary tasks based on the decomposition strategy and propose a multi-task framework for MNER. Specifically, we first decompose MNER into two auxiliary tasks: entity boundary detection task and entity category classification task . Here, the former treats only the text modality as input and outputs the boundary labels, since it can achieve satisfactory boundary results by itself. The latter uses two modalities to yield category labels where image modality is dedicated to disambiguating categories. These two auxiliary tasks allow the effective exploitation of text and image modalities and put them back into their respective roles. Then, we vectorize their results to improve entity recognition using label clues from auxiliary tasks. Finally, we fuse features from text and image modalities and label embeddings from auxiliary tasks to fulfill MNER. Experimental results on two widely used MNER datasets show that our framework can yield new SOTA performance.

Multimodal aspect-based sentiment analysis (MABSA) aims to determine the sentiment polarity of each aspect mentioned in the text based on multimodal content. Various approaches have been proposed to model multimodal sentiment features for each aspect via modal interactions. However, most existing approaches have two shortcomings: (1) The representation gap between textual and visual modalities may increase the risk of misalignment in modal interactions; (2) In some examples where the image is not related to the text, the visual information may not enrich the textual modality when learning aspect-based sentiment features. In such cases, blindly leveraging information from visual modal may introduce noises in reasoning the aspect-based sentiment expressions. To tackle the shortcomings mentioned above, we propose an end-to-end MABSA framework with image conversion and noise filtration. Specifically, to bridge the representation gap in different modalities, we attempt to translate images into the input space of a pre-trained language model (PLM). To this end, we develop an image-to-text conversion module that can convert an image to an implicit sequence of token embedding. Moreover, an aspect-oriented filtration module is devised to alleviate the noise in the implicit token embeddings, which consists of two attention operations. The former aims to create an enhanced aspect embedding as a query, and the latter seeks to use this query to retrieve relevant auxiliary information from the implicit token embeddings to supplement the textual content. After filtering the noise, we leverage a PLM to encode the text, aspect, and image prompt derived from filtered implicit token embeddings as sentiment features to perform aspect-based sentiment prediction. Experimental results on two MABSA datasets show that our framework achieves state-of-the-art performance. Furthermore, extensive experimental analysis demonstrates the proposed framework has superior robustness and efficiency.

Meeting summarization has emerged as a promising technique for providing users with condensed summaries. However, existing work has focused on training models on centralized data, neglecting real-world scenarios where meeting data are infeasible to collect centrally, due to their sensitive nature. This gap motivates us to explore federated learning for meeting summarization. Two critical challenges impede progress. First, state-of-the-art summarizers are based on parameter-heavy pre-trained models. Exchanging such a model's parameters across clients imposes large bandwidth costs. Second, as real-world meeting data belong to various domains and are distributed across clients, they are instances of non-identically and independently distributed (non-IID). IID assumptions do not hold, which changes which forms of learning algorithms best apply. To address this, we proposeAdapter-based Federated Selective Knowledge Distillation(AdaFedSelecKD) for training performant client models. Specifically, we develop an adapter-based summarization model where two adapters cooperatively facilitate learning using fewer parameters to reduce communication costs. Then, we devise a selective knowledge distillation strategy, assisting clients in robustly handling domain-focused modelling on their own data, while leveraging global parameters based on non-IID data. Extensive experiments on the QMSum benchmark demonstrateAdaFedSelecKDcan achieve comparable performance with powerful centralized training methods, and shows its generalizability and robustness.

Given data with label noise (i.e., incorrect data), deep neural networks would gradually memorize the label noise and impair model performance. To relieve this issue, curriculum learning is proposed to improve model performance and generalization by ordering training samples in a meaningful (e.g., easy to hard) sequence. Previous work takes incorrect samples as generic hard ones without discriminating between hard samples (i.e., hard samples in correct data) and incorrect samples. Indeed, a model should learn from hard samples to promote generalization rather than overfit to incorrect ones. In this article, we address this problem by appending a novel loss functionDiscrimLoss, on top of the existing task loss. Its main effect is to automatically and stably estimate the importance of easy samples and difficult samples (including hard and incorrect samples) at the early stages of training to improve the model performance. Then, during the following stages, DiscrimLoss is dedicated to discriminating between hard and incorrect samples to improve the model generalization. Such a training strategy can be formulated dynamically in a self-supervised manner, effectively mimicking the main principle of curriculum learning. Experiments on image classification, image regression, text sequence regression, and event relation reasoning demonstrate the versatility and effectiveness of our method, particularly in the presence of diversified noise levels.

Understanding temporal commonsense concepts, such as times of occurrence and durations is crucial for event-centric language understanding. Reasoning about such temporal concepts in a complex context requires reasoning over both the stated context and the world knowledge that underlines it. A recent study shows massive pre-trained LM still struggle with such temporal reasoning under complex contexts (e.g., dialog) because they only implicitly encode the relevant contexts and fail to explicitly uncover the underlying logical compositions for complex inference, thus may not be robust enough. In this work, we propose to augment LMs with the temporal logic induction ability, which frames the temporal reasoning by defining three modular components: temporal dependency inducer and temporal concept defuzzifier and logic validator. The former two components disentangle the explicit/implicit dependency between temporal concepts across context (before, after, ...) and the specific meaning of fuzzy temporal concepts, respectively, while the validator combines the intermediate reasoning clues for robust contextual reasoning about the temporal concepts. Extensive experimental results on TIMEDIAL, a challenging dataset for temporal reasoning over dialog, show that our method, Logic Induction Enhanced Contextualized TEmporal Reasoning (LECTER), can yield great improvements over the traditional language model for temporal reasoning.

Causal Emotion Entailment aims to identify causal utterances that are responsible for the target utterance with a non-neutral emotion in conversations. Previous works are limited in thorough understanding of the conversational context and accurate reasoning of the emotion cause. To this end, we propose Knowledge-Bridged Causal Interaction Network (KBCIN) with commonsense knowledge (CSK) leveraged as three bridges. Specifically, we construct a conversational graph for each conversation and leverage the event-centered CSK as the semantics-level bridge (S-bridge) to capture the deep inter-utterance dependencies in the conversational context via the CSK-Enhanced Graph Attention module. Moreover, social-interaction CSK serves as emotion-level bridge (E-bridge) and action-level bridge (A-bridge) to connect candidate utterances with the target one, which provides explicit causal clues for the Emotional Interaction module and Actional Interaction module to reason the target emotion. Experimental results show that our model achieves better performance over most baseline models. Our source code is publicly available at https://github.com/circle-hit/KBCIN.

Although large-scale video-language pre-training models, which usually build a global alignment between the video and the text, have achieved remarkable progress on various downstream tasks, the idea of adopting fine-grained information during the pre-training stage is not well explored. In this work, we propose STOA-VLP, a pre-training framework that jointly models object and action information across spatial and temporal dimensions. More specifically, the model regards object trajectories across frames and multiple action features from the video as fine-grained features. Besides, We design two auxiliary tasks to better incorporate both kinds of information into the pre-training process of the video-language model. The first is the dynamic object-text alignment task, which builds a better connection between object trajectories and the relevant noun tokens. The second is the spatial-temporal action set prediction, which guides the model to generate consistent action features by predicting actions found in the text. Extensive experiments on three downstream tasks (video captioning, text-video retrieval, and video question answering) demonstrate the effectiveness of our proposed STOA-VLP (e.g. 3.7 Rouge-L improvements on MSR-VTT video captioning benchmark, 2.9% accuracy improvements on MSVD video question answering benchmark, compared to previous approaches).

Although achieving promising performance, current Natural Language Understanding models tend to utilize dataset biases instead of learning the intended task, which always leads to performance degradation on out-of-distribution (OOD) samples.To increase the performance stability, previous debiasing methods empirically capture bias features from data to prevent the model from corresponding biases.However, our analyses show that the empirical debiasing methods may fail to capture part of the dataset biases and mistake semantic information of input text as biases, which limits the effectiveness of debiasing.To address these issues, we propose a debiasing framework IEGDB that comprehensively detects the dataset biases to induce a set of biased features, and purify the biased features with the guidance of information entropy.Experimental results show that IEGDB can consistently improve the stability of performance on OOD datasets for a set of widely adopted NLU models.

Yuxuan Gu, Xiaocheng Feng, Sicheng Ma, Lingyuan Zhang, Heng Gong, Weihong Zhong, Bing Qin. Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2023.

Multilingual neural machine translation has witnessed remarkable progress in recent years.However, the long-tailed distribution of multilingual corpora poses a challenge of Pareto optimization, i.e., optimizing for some languages may come at the cost of degrading the performance of others.Existing balancing training strategies are equivalent to a series of Pareto optimal solutions, which trade off on a Pareto frontier 1 .In this work, we propose a new training framework, Pareto Mutual Distillation (Pareto-MD), towards pushing the Pareto frontier outwards rather than making trade-offs.Specifically, Pareto-MD collaboratively trains two Pareto optimal solutions that favor different languages and allows them to learn from the strengths of each other via knowledge distillation.Furthermore, we introduce a novel strategy to enable stronger communication between Pareto optimal solutions and broaden the applicability of our approach.Experimental results on the widely-used WMT and TED datasets show that our method significantly pushes the Pareto frontier and outperforms baselines by up to +2.46 BLEU 2 .

Decoding text stimuli from cognitive signals (e.g.fMRI) enhances our understanding of the human language system, paving the way for building versatile Brain-Computer Interface.However, existing studies largely focus on decoding individual word-level fMRI volumes from a restricted vocabulary, which is far too idealized for real-world application.In this paper, we propose fMRI2text, the first openvocabulary task aiming to bridge fMRI time series and human language.Furthermore, to explore the potential of this new task, we present a baseline solution, UniCoRN: the Unified Cognitive Signal ReconstructioN for Brain Decoding.By reconstructing both individual time points and time series, UniCoRN establishes a robust encoder for cognitive signals (fMRI & EEG).Leveraging a pre-trained language model as decoder, UniCoRN proves its efficacy in decoding coherent text from fMRI series across various split settings.Our model achieves a 34.77%BLEU score on fMRI2text, and a 37.04% BLEU when generalized to EEGto-text decoding, thereby surpassing the former baseline.Experimental results indicate the feasibility of decoding consecutive fMRI volumes, and the effectiveness of decoding different cognitive signals using a unified structure.

Adverse drug-drug interactions (DDIs) can compromise the effectiveness of concurrent drug administration, posing a significant challenge in healthcare.As the development of new drugs continues, the potential for unknown adverse effects resulting from DDIs becomes a growing concern.Traditional computational methods for DDI prediction may fail to capture interactions for new drugs due to the lack of knowledge.In this paper, we introduce a new problem setup as zero-shot DDI prediction that deals with the case of new drugs.Leveraging textual information from online databases like DrugBank and PubChem, we propose an innovative approach TextDDI with a language model-based DDI predictor and a reinforcement learning (RL)-based information selector, enabling the selection of concise and pertinent text for accurate DDI prediction on new drugs.Empirical results show the benefits of the proposed approach on several settings including zero-shot and few-shot DDI prediction, and the selected texts are semantically relevant.Our code and data are available at https: //github.com/zhufq00/DDIs-Prediction.

Action item detection aims at recognizing sentences containing information about actionable tasks, which can help people quickly grasp core tasks in the meeting without going through the redundant meeting contents. Therefore, in this paper, we thoroughly describe our carefully designed solution for the Action Item Detection Track of the General Meeting Understanding and Generation (MUG) challenge in the ICASSP 2023 Signal Processing Grand Challenge. Specifically, we systematically analyze the task instances provided by MUG and find that the key ingredient for successful action item detection is leveraging the dialogue context information into consideration. To this end, we design a simple and effective method for modelling context and utterance information concurrently. The experimental results show our method achieves remarkable improvements over baseline models, with an absolute increase of 0.62 of the F1score on the validation set. The stable generalizability of our method is further verified by our score on the final test set1.

Emotion Forecasting in Conversations (EFC), the task aims to predict the emotion of next utterance (yet to come), has received more and more attention in recent years. However, this task ignores the one-to-many feature of dialogue and its prediction target is emotion label, which is flawed in most cases. In this work, we propose a new task: Emotion Distribution Forecasting in Conversations (EDFC), which aims to predict the emotion distribution of next utterance. Although this task is more reasonable in real applications, it can only learn using emotion labels in most cases because of the difficulty in obtaining emotion distribution. To address it, we explore the positive role of topic in this task and propose a topic-enhanced approach. Specifically, we first obtain the topic-based emotion distribution prior through topic model and emotion generation model, and then use the emotion distribution prior to enhance original label learning model. To effectively evaluate the distribution prediction results, we construct two datasets for this task, and the experimental results prove the feasibility of the EDFC task as well as the effectiveness of our approach.

Recent research has revealed that deep neural networks often take dataset biases as a shortcut to make decisions rather than understand tasks, leading to failures in real-world applications. In this study, we focus on the spurious correlation between word features and labels that models learn from the biased data distribution of training data. In particular, we define the word highly co-occurring with a specific label as biased word, and the example containing biased word as biased example. Our analysis shows that biased examples are easier for models to learn, while at the time of prediction, biased words make a significantly higher contribution to the models' predictions, and models tend to assign predicted labels over-relying on the spurious correlation between words and labels. To mitigate models' over-reliance on the shortcut (i.e. spurious correlation), we propose a training strategy Less-Learn-Shortcut (LLS): our strategy quantifies the biased degree of the biased examples and down-weights them accordingly. Experimental results on Question Matching, Natural Language Inference and Sentiment Analysis tasks show that LLS is a task-agnostic strategy and can improve the model performance on adversarial data while maintaining good performance on in-domain data.

Dynamic scene graph generation aims to identify visual relationships (subject-predicate-object) in frames based on spatio-temporal contextual information in the video. Previous work implicitly models the spatio-temporal interaction simultaneously, which leads to entanglement of spatio-temporal contextual information. To this end, we propose a Grafting-Then-Reassembling framework (GTR), which explicitly extracts intra-frame spatial information and inter-frame temporal information in two separate stages to decouple spatio-temporal contextual information. Specifically, we first graft a static scene graph generation model to generate static visual relationships within frames. Then we propose the temporal dependency model to extract the temporal dependencies across frames, and explicitly reassemble static visual relationships into dynamic scene graphs. Experimental results show that GTR achieves the state-of-the-art performance on Action Genome dataset. Further analyses reveal that the reassembling stage is crucial to the success of our framework.

Modern Entity Linking (EL) systems entrench a popularity bias, yet there is no dataset focusing on tail and emerging entities in languages other than English. We present Hansel, a new benchmark in Chinese that fills the vacancy of non-English few-shot and zero-shot EL challenges. The test set of Hansel is human annotated and reviewed, created with a novel method for collecting zero-shot EL datasets. It covers 10K diverse documents in news, social media posts and other web articles, with Wikidata as its target Knowledge Base. We demonstrate that the existing state-of-the-art EL system performs poorly on Hansel ([email protected] of 36.6% on Few-Shot). We then establish a strong baseline that scores a [email protected] of 46.2% on Few-Shot and 76.6% on Zero-Shot on our dataset. We also show that our baseline achieves competitive results on TAC-KBP2015 Chinese Entity Linking task. Datasets and codes are released at https://github.com/HITsz-TMG/Hansel.

Understanding causality has vital importance for various Natural Language Processing (NLP) applications.Beyond the labeled instances, conceptual explanations of the causality can provide deep understanding of the causal facts to facilitate the causal reasoning process.However, such explanation information still remains absent in existing causal reasoning resources.In this paper, we fill this gap by presenting a human-annotated explainable CAusal REasoning dataset (e-CARE), which contains over 21K causal reasoning questions, together with natural language formed explanations of the causal questions.Experimental results show that generating valid explanations for causal facts still remains especially challenging for the state-of-the-art models, and the explanation information can be helpful for promoting the accuracy and stability of causal reasoning models.

We study the problem of integrating cognitive language processing signals (e.g., eye-tracking or EEG data) into pre-trained language models like BERT. Existing methods typically fine-tune pre-trained models on cognitive data, ignoring the semantic gap between the texts and cognitive signals. To fill the gap, we propose CogBERT, a framework that can induce fine-grained cognitive features from cognitive data and incorporate cognitive features into BERT by adaptively adjusting the weight of cognitive features for different NLP tasks. Extensive experiments show that: (1) Cognition-guided pre-trained models can consistently perform better than basic pre-trained models on ten NLP tasks. (2) Different cognitive features contribute differently to different NLP tasks. Based on this observation, we give a fine-grained explanation of why cognitive data is helpful for NLP. (3) Different transformer layers of pre-trained models should encode different cognitive features, with word-level cognitive features at the bottom and semantic-level cognitive features at the top. (4) Attention visualization demonstrates that CogBERT aligns with human gaze patterns and improves its natural language comprehension ability.

Prompt-based fine-tuning for pre-trained models has proven effective for many natural language processing tasks under few-shot settings in general domain. However, tuning with prompt in biomedical domain has not been investigated thoroughly. Biomedical words are often rare in general domain, but quite ubiquitous in biomedical contexts, which dramatically deteriorates the performance of pre-trained models on downstream biomedical applications even after fine-tuning, especially in low-resource scenarios. We propose a simple yet effective approach to helping models learn rare biomedical words during tuning with prompt. Experimental results show that our method can achieve up to 6% improvement in biomedical natural language inference task without any extra parameters or training steps using few-shot vanilla prompt settings.

Dataset bias in stance detection tasks allows models to achieve superior performance without using targets. Most existing debiasing methods are task-agnostic, which fail to utilize task knowledge to better discriminate between genuine and bias features. Motivated by how humans tackle stance detection tasks, we propose to incorporate the stance reasoning process as task knowledge to assist in learning genuine features and reducing reliance on bias features. The full stance reasoning process usually involves identifying the span of the mentioned target and corresponding opinion expressions, such fine-grained annotations are hard and expensive to obtain. To alleviate this, we simplify the stance reasoning process to relax the granularity of annotations from token-level to sentence-level, where labels for sub-tasks can be easily inferred from existing resources. We further implement those sub-tasks by maximizing mutual information between the texts and the opinioned targets. To evaluate whether stance detection models truly understand the task from various aspects, we collect and construct a series of new test sets. Our proposed model achieves better performance than previous task-agnostic debiasing methods on most of those new test sets while maintaining comparable performances to existing stance detection models.

As an emerging research topic in natural language processing community, emotion recognition in multi-party conversations has attained increasing interest. Previous approaches that focus either on dyadic or multi-party scenarios exert much effort to cope with the challenge of emotional dynamics and achieve appealing results. However, since emotional interactions among speakers are often more complicated within the entangled multi-party conversations, these works are limited in capturing effective emotional clues in conversational context. In this work, we propose Mutual Conversational Detachment Network (MuCDN) to clearly and effectively understand the conversational context by separating conversations into detached threads. Specifically, two detachment ways are devised to perform context and speaker-specific modeling within detached threads and they are bridged through a mutual module. Experimental results on two datasets show that our model achieves better performance over the baseline models.

Argument mining (AM) is a challenging task as it requires recognizing the complex argumentation structures involving multiple subtasks.To handle all subtasks of AM in an end-to-end fashion, previous works generally transform AM into a dependency parsing task.However, such methods largely require complex pre-and post-processing to realize the task transformation.In this paper, we investigate the endto-end AM task from a novel perspective by proposing a generative framework, in which the expected outputs of AM are framed as a simple target sequence.Then, we employ a pretrained sequence-to-sequence language model with a constrained pointer mechanism (CPM) to model the clues for all the subtasks of AM in the light of the target sequence.Furthermore, we devise a reconstructed positional encoding (RPE) to alleviate the order biases induced by the autoregressive generation paradigm.Experimental results show that our proposed framework achieves new state-of-the-art performance on two AM benchmarks. 1

Multi-aspect controllable text generation is a more challenging and practical task than single-aspect control. Existing methods achieve complex multi-aspect control by fusing multiple controllers learned from single-aspect, but suffer from attribute degeneration caused by the mutual interference of these controllers. To address this, we provide observations on attribute fusion from a distributional perspective and propose to directly search for the intersection areas of multiple attribute distributions as their combination for generation. Our method first estimates the attribute space with an autoencoder structure. Afterward, we iteratively approach the intersections by jointly minimizing distances to points representing different attributes. Finally, we map them to attribute-relevant sentences with a prefix-tuning-based decoder. Experiments on the three-aspect control task, including sentiment, topic, and detoxification aspects, reveal that our method outperforms several strong baselines on attribute relevance and text quality and achieves the SOTA. Further analysis also supplies some explanatory support for the effectiveness of our approach.

Although all-in-one-model multilingual neural machine translation (MNMT) has achieved remarkable progress, the convergence inconsistency in the joint training is ignored, i.e.,different language pairs reaching convergence in different epochs.This leads to the trained MNMT model over-fitting low-resource language translations while under-fitting highresource ones.In this paper, we propose a novel training strategy named LSSD (Language-Specific Self-Distillation), which can alleviate the convergence inconsistency and help MNMT models achieve the best performance on each language pair simultaneously.Specifically, LSSD picks up language-specific best checkpoints for each language pair to teach the current model on the fly.Furthermore, we systematically explore three sample-level manipulations of knowledge transferring.Experimental results on three datasets show that LSSD obtains consistent improvements towards all language pairs and achieves the state-of-the-art 1 .

On vision-language understanding (VLU) tasks, fusion-encoder vision-language models achieve superior results but sacrifice efficiency because of the simultaneous encoding of images and text.On the contrary, the dual-encoder model that separately encodes images and text has the advantage in efficiency, while failing on VLU tasks due to the lack of deep cross-modal interactions.To get the best of both worlds, we propose DIDE 1 , a framework that distills the knowledge of the fusion-encoder teacher model into the dual-encoder student model.Since the cross-modal interaction is the key to the superior performance of teacher model but is absent in the student model, we encourage the student not only to mimic the predictions of teacher, but also to calculate the cross-modal attention distributions and align with the teacher.Experimental results demonstrate that DIDE is competitive with the fusion-encoder teacher model in performance (only a 1% drop) while enjoying 4× faster inference.Further analyses reveal that the proposed cross-modal attention distillation is crucial to the success of our framework.

Noisy labels are ubiquitous in natural language processing (NLP) tasks.Existing work, namely learning with noisy labels in NLP, is often limited to dedicated tasks or specific training procedures, making it hard to be widely used.To address this issue, SGD noise has been explored to provide a more general way to alleviate the effect of noisy labels by involving benign noise in the process of stochastic gradient descent.However, previous studies exert identical perturbation for all samples, which may cause overfitting on incorrect ones or optimizing correct ones inadequately.To facilitate this, we propose a novel stochastic tailor-made gradient noise (STGN), mitigating the effect of inherent label noise by introducing tailor-made benign noise for each sample.Specifically, we investigate multiple principles to precisely and stably discriminate correct samples from incorrect ones and thus apply different intensities of perturbation to them.A detailed theoretical analysis shows that STGN has good properties, beneficial for model generalization.Experiments on three different NLP tasks demonstrate the effectiveness and versatility of STGN.Also, STGN can boost existing robust training methods. 1

Causal chain reasoning (CCR) is an essential ability for many decision-making AI systems, which requires the model to build reliable causal chains by connecting causal pairs.However, CCR suffers from two main transitive problems: threshold effect and scene drift.In other words, the causal pairs to be spliced may have a conflicting threshold boundary or scenario.To address these issues, we propose a novel Reliable Causal chain reasoning framework (ReCo), which introduces exogenous variables to represent the threshold and scene factors of each causal pair within the causal chain, and estimates the threshold and scene contradictions across exogenous variables via structural causal recurrent neural networks (SRNN).Experiments show that ReCo outperforms a series of strong baselines on both Chinese and English CCR datasets.Moreover, by injecting reliable causal chain knowledge distilled by ReCo, BERT can achieve better performances on four downstream causal-related tasks than BERT models enhanced by other kinds of knowledge.

Aspect-level multimodal sentiment analysis, which aims to identify the sentiment of the target aspect from multimodal data, recently has attracted extensive attention in the community of multimedia and natural language processing.Despite the recent success in textual aspect-based sentiment analysis, existing models mainly focused on utilizing the object-level semantic information in the image but ignore explicitly using the visual emotional cues, especially the facial emotions.How to distill visual emotional cues and align them with the textual content remains a key challenge to solve the problem.In this work, we introduce a facesensitive image-to-emotional-text translation (FITE) method, which focuses on capturing visual sentiment cues through facial expressions and selectively matching and fusing with the target aspect in textual modality.To the best of our knowledge, we are the first that explicitly utilize the emotional information from images in the multimodal aspect-based sentiment analysis task.Experiment results show that our method achieves state-of-the-art results on the Twitter-2015 and Twitter-2017 datasets.The improvement demonstrates the superiority of our model in capturing aspect-level sentiment in multimodal data with facial expressions 1 .

Dialogue summarization aims to condense the original dialogue into a shorter version covering salient information, which is a crucial way to reduce dialogue data overload. Recently, the promising achievements in both dialogue systems and natural language generation techniques drastically lead this task to a new landscape, which results in significant research attentions. However, there still remains a lack of a comprehensive survey for this task. To this end, we take the first step and present a thorough review of this research field carefully and widely. In detail, we systematically organize the current works according to the characteristics of each domain, covering meeting, chat, email thread, customer service and medical dialogue. Additionally, we provide an overview of publicly available research datasets as well as organize two leaderboards under unified metrics. Furthermore, we discuss some future directions, including faithfulness, multi-modal, multi-domain and multi-lingual dialogue summarization, and give our thoughts respectively. We hope that this first survey of dialogue summarization can provide the community with a quick access and a general picture to this task and motivate future researches.

Existing approaches for sarcasm detection are mainly based on supervised learning, in which the promising performance largely depends on a considerable amount of labeled data or extra information. In the real world scenario, however, the abundant labeled data or extra information requires high labor cost, not to mention that sufficient annotated data is unavailable in many low-resource conditions. To alleviate this dilemma, we investigate sarcasm detection from an unsupervised perspective, in which we explore a masking and generation paradigm in the context to extract the context incongruities for learning sarcastic expression. Further, to improve the feature representations of the sentences, we use unsupervised contrastive learning to improve the sentence representation based on the standard dropout. Experimental results on six perceived sarcasm detection benchmark datasets show that our approach outperforms baselines. Simultaneously, our unsupervised method obtains comparative performance with supervised methods for the intended sarcasm dataset.

MOTIVATION: With the rapid development of precision medicine, a large amount of health data (such as electronic health records, gene sequencing, medical images, etc.) has been produced. It encourages more and more interest in data-driven insight discovery from these data. A reasonable way to verify the derived insights is by checking evidence from biomedical literature. However, manual verification is inefficient and not scalable. Therefore, an intelligent technique is necessary to solve this problem. RESULTS: This article introduces a framework for biomedical evidence engineering, addressing this problem more effectively. The framework consists of a biomedical literature retrieval module and an evidence extraction module. The retrieval module ensembles several methods and achieves state-of-the-art performance in biomedical literature retrieval. A BERT-based evidence extraction model is proposed to extract evidence from literature in response to queries. Moreover, we create a dataset with 1 million examples of biomedical evidence, 10 000 of which are manually annotated. AVAILABILITY AND IMPLEMENTATION: Datasets are available at https://github.com/SendongZhao.

Owing to the availability of various large-scale Machine Reading Comprehension ( MRC ) datasets, building an effective model to extract passage spans for question answering has been well studied in previous works. However, in reality, there are some questions that cannot be answered through the passage information, which brings more challenges to this task. In this article, we propose an Interactive Gated Decoder ( IG Decoder ), which focuses on modeling the interactions between the answer span prediction and no-answer prediction with a gating mechanism. We also propose a simple but effective approach for automatically generating pseudo training data, which aims to enrich the training data of the unanswerable questions. Experimental results on popular benchmark SQuAD 2.0 and NewsQA show that the proposed approaches yield consistent improvements over traditional BERT-large and strong ALBERT-xxlarge baseline systems. We also provide detailed ablations of the proposed method and error analysis on hard samples, which could be helpful in future research.

The limited size of existing query-focused summarization datasets renders training data-driven summarization models challenging. Meanwhile, the manual construction of a query-focused summarization corpus is costly and time-consuming. In this paper, we use Wikipedia to automatically collect a large query-focused summarization dataset (named WikiRef) of more than 280,000 examples, which can serve as a means of data augmentation. We also develop a BERT-based query-focused summarization model (Q-BERT) to extract sentences from the documents as summaries. To better adapt a huge model containing millions of parameters to tiny benchmarks, we identify and fine-tune only a sparse subnetwork, which corresponds to a small fraction of the whole model parameters. Experimental results on three DUC benchmarks show that the model pre-trained on WikiRefhas already achieved reasonable performance. After fine-tuning on the specific benchmark datasets, the model with data augmentation outperforms strong comparison systems. Moreover, both our proposed Q-BERT model and subnetwork fine-tuning further improve the model performance.

Online medical consult and offline medical check-in have generated a large amount of health-related data in medical forums and patient records. However, exploiting the user-generated content for orienting patients online and assisting medical checkup offline is nontrivial due to the sparsity of symptom-disease associations. The serious sparsity is caused by the informal/chatty expressions of symptoms in the data.

Li Du, Xiao Ding, Ting Liu, Bing Qin. Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers). 2021.

Li Du, Xiao Ding, Kai Xiong, Ting Liu, Bing Qin. Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers). 2021.

Xiachong Feng, Xiaocheng Feng, Libo Qin, Bing Qin, Ting Liu. Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers). 2021.

Providing timely accessibility reminders (such as closed and relocated) of a point-of-interest (POI) plays a vital role in improving user satisfaction of finding places and making visiting decisions. However, it is difficult to keep the POI database in sync with the real-world counterparts due to the dynamic nature of business changes and innovations. To alleviate this problem, we formulate and present a practical solution that jointly extracts POI mentions and identifies their coupled accessibility labels from unstructured text (hereafter referred to as joint POI and accessibility extraction). We approach this task as a sequence tagging problem, where the goal is to produce (POI name, accessibility label) pairs from unstructured text. This task is challenging because of two main issues: (1) POI names are often newly-coined words so as to successfully register new entities or brands and (2) there may exist multiple pairs in the text, which necessitates dealing with one-to-many or many-to-one mapping to make each POI coupled with its matching accessibility label. To this end, we propose a Geographic-Enhanced and Dependency-guIded sequence Tagging (GEDIT) model to concurrently address the two challenges. First, to alleviate challenge #1, we develop a geographic-enhanced pre-trained model to learn the text representations, which is able to significantly relieve the problem of newly-coined words. Second, to mitigate challenge #2, we apply a relational graph convolutional network to learn the tree node representations from the parsed dependency tree, which enables us to establish a correlation between a POI and its accessibility label. Finally, we construct a neural sequence tagging model by integrating and feeding the previously pre-learned representations into a CRF layer. Extensive experiments conducted on a real-world dataset demonstrate the superiority and effectiveness of GEDIT. In addition, it has already been deployed in production at Baidu Maps, and it successfully keeps processing hundreds of thousands of Web documents every week. Statistics show that the proposed solution can save significant human effort and labor costs to deal with the same amount of documents, which confirms that it is a practical way for POI accessibility maintenance.

Non-autoregressive neural machine translation, which decomposes the dependence on previous target tokens from the inputs of the decoder, has achieved impressive inference speedup but at the cost of inferior accuracy.Previous works employ iterative decoding to improve the translation by applying multiple refinement iterations.However, a serious drawback is that these approaches expose the serious weakness in recognizing the erroneous translation pieces.In this paper, we propose an architecture named REWRITENAT to explicitly learn to rewrite the erroneous translation pieces.Specifically, REWRITENAT utilizes a locator module to locate the erroneous ones, which are then revised into the correct ones by a revisor module.Towards keeping the consistency of data distribution with iterative decoding, an iterative training strategy is employed to further improve the capacity of rewriting.Extensive experiments conducted on several widely-used benchmarks show that REWRITE-NAT can achieve better performance while significantly reducing decoding time, compared with previous iterative decoding strategies.In particular, REWRITENAT can obtain competitive results with autoregressive translation on WMT14 En↔De, En→Fr and WMT16 Ro→En translation benchmarks 1 .

Many open-domain question answering problems can be cast as a textual entailment task, where a question and candidate answers are concatenated to form hypotheses.A QA system then determines if the supporting knowledge bases, regarded as potential premises, entail the hypotheses.In this paper, we investigate a neural-symbolic QA approach that integrates natural logic reasoning within deep learning architectures, towards developing effective and yet explainable question answering models.The proposed model gradually bridges a hypothesis and candidate premises following natural logic inference steps to build proof paths.Entailment scores between the acquired intermediate hypotheses and candidate premises are measured to determine if a premise entails the hypothesis.As the natural logic reasoning process forms a tree-like, hierarchical structure, we embed hypotheses and premises in a Hyperbolic space rather than Euclidean space to acquire more precise representations.Empirically, our method outperforms prior work on answering multiple-choice science questions, achieving the best results on two publicly available datasets.The natural logic inference process inherently provides evidence to help explain the prediction process.

Meeting summarization is a challenging task due to its dynamic interaction nature among multiple speakers and lack of sufficient training data. Existing methods view the meeting as a linear sequence of utterances while ignoring the diverse relations between each utterance. Besides, the limited labeled data further hinders the ability of data-hungry neural models. In this paper, we try to mitigate the above challenges by introducing dialogue-discourse relations. First, we present a Dialogue Discourse-Dware Meeting Summarizer (DDAMS) to explicitly model the interaction between utterances in a meeting by modeling different discourse relations. The core module is a relational graph encoder, where the utterances and discourse relations are modeled in a graph interaction manner. Moreover, we devise a Dialogue Discourse-Aware Data Augmentation (DDADA) strategy to construct a pseudo-summarization corpus from existing input meetings, which is 20 times larger than the original dataset and can be used to pretrain DDAMS. Experimental results on AMI and ICSI meeting datasets show that our full system can achieve SOTA performance. Our codes and outputs are available at https://github.com/xcfcode/DDAMS/.

Multiparty Dialogue Machine Reading Comprehension (MRC) differs from traditional MRC as models must handle the complex dialogue discourse structure, previously unconsidered in traditional MRC. To fully exploit such discourse structure in multiparty dialogue, we present a discourse-aware dialogue graph neural network, DADgraph, which explicitly constructs the dialogue graph using discourse dependency links and discourse relations. To validate our model, we perform experiments on the Molweni corpus, a large-scale MRC dataset built over multiparty dialogue annotated with discourse structure. Experiments on Molweni show that our discourse-aware model achieves statistically significant improvements compared against strong neural network MRC baselines.

Bidirectional Encoder Representations from Transformers (BERT) has shown marvelous improvements across various NLP tasks, and its consecutive variants have been proposed to further improve the performance of the pre-trained language models. In this paper, we aim to first introduce the whole word masking (wwm) strategy for Chinese BERT, along with a series of Chinese pre-trained language models. Then we also propose a simple but effective model called MacBERT, which improves upon RoBERTa in several ways. Especially, we propose a new masking strategy called MLM as correction (Mac). To demonstrate the effectiveness of these models, we create a series of Chinese pre-trained language models as our baselines, including BERT, RoBERTa, ELECTRA, RBT, etc. We carried out extensive experiments on ten Chinese NLP tasks to evaluate the created Chinese pre-trained language models as well as the proposed MacBERT. Experimental results show that MacBERT could achieve state-of-the-art performances on many NLP tasks, and we also ablate details with several findings that may help future research. We open-source our pre-trained language models for further facilitating our research community.

Conditions play an essential role in biomedical statements. However, existing biomedical knowledge graphs (BioKGs) only focus on factual knowledge, organized as a flat relational network of biomedical concepts. These BioKGs ignore the conditions of the facts being valid, which loses essential contexts for knowledge exploration and inference. We consider both facts and their conditions in biomedical statements and proposed a three-layered information-lossless representation of BioKG. The first layer has biomedical concept nodes, attribute nodes. The second layer represents both biomedical fact and condition tuples by nodes of the relation phrases, connecting to the subject and object in the first layer. The third layer has nodes of statements connecting to a set of fact tuples and/or condition tuples in the second layer. We transform the BioKG construction problem into a sequence labeling problem based on a novel designed tag schema. We design a Multi-Input Multi-Output sequence labeling model (MIMO) that learns from multiple input signals and generates proper number of multiple output sequences for tuple extraction. Experiments on a newly constructed dataset show that MIMO outperforms the existing methods. Further case study demonstrates that the BioKGs constructed provide a good understanding of the biomedical statements.

Neural semantic parsing has achieved impressive results in recent years, yet its success relies on the availability of large amounts of supervised data. Our goal is to learn a neural semantic parser when only prior knowledge about a limited number of simple rules is available, without access to either annotated programs or execution results. Our approach is initialized by rules, and improved in a back-translation paradigm using generated question-program pairs from the semantic parser and the question generator. A phrase table with frequent mapping patterns is automatically derived, also updated as training progresses, to measure the quality of generated instances. We train the model with model-agnostic meta-learning to guarantee the accuracy and stability on examples covered by rules, and meanwhile acquire the versatility to generalize well on examples uncovered by rules. Results on three benchmark datasets with different domains and programs show that our approach incrementally improves the accuracy. On WikiSQL, our best model is comparable to the state-of-the-art system learned from denotations.

Self-attention networks (SANs) with selective mechanism has produced substantial improvements in various NLP tasks by concentrating on a subset of input words.However, the underlying reasons for their strong performance have not been well explained.In this paper, we bridge the gap by assessing the strengths of selective SANs (SSANs), which are implemented with a flexible and universal Gumbel-Softmax.Experimental results on several representative NLP tasks, including natural language inference, semantic role labelling, and machine translation, show that SSANs consistently outperform the standard SANs.Through well-designed probing experiments, we empirically validate that the improvement of SSANs can be attributed in part to mitigating two commonly-cited weaknesses of SANs: word order encoding and structure modeling.Specifically, the selective mechanism improves SANs by paying more attention to content words that contribute to the meaning of the sentence.The code and data are released at https://github.com/xwgeng/SSAN.

Knowledge plays an essential role in biomedical literature search (BLS) systems, filling the semantic gap between queries and documents. Knowledge bases, constructed by human experts or machine learning methods, are generally regarded as the main sources serving external knowledge. However, a good knowledge base must balances its particularity and generalization, resulting limited knowledge coverage and utilization to BLS systems. Considering massive documents in a BLS system, and recently developing Open IE techniques by which we can automatically extract structured knowledge from documents, how about harnessing distilled internal knowledge rather than external knowledge to conduct BLS systems? Internal knowledge, providing tailored particular knowledge to BLS systems, is supposed to lead better knowledge utilization and much more competitive performance on literature search. In this paper, we design an novel internal knowledge driven BLS system upon a Multi-layered Encoders incorporating Multi-layered internal Knowledge graph, called MEMK. MEMK harnesses distilled internal structural knowledge, empowering interactive representations learning of query and documents. The experiments show that MEMK outperforms strong baselines on a public benchmark, and internal knowledge based query expansion can further improve the performance to a new state of the art.

Although neural table-to-text models have achieved remarkable progress with the help of largescale datasets, they suffer insufficient learning problem with limited training data.Recently, pretrained language models show potential in few-shot learning with linguistic knowledge learnt from pretraining on large-scale corpus.However, benefiting table-to-text generation in few-shot setting with the powerful pretrained language model faces three challenges, including (1) the gap between the task's structured input and the natural language input for pretraining language model.(2) The lack of modeling for table structure and ( 3) improving text fidelity with less incorrect expressions that are contradicting to the table.To address aforementioned problems, we propose TableGPT for table-to-text generation.At first, we utilize table transformation module with template to rewrite structured table in natural language as input for GPT-2.In addition, we exploit multi-task learning with two auxiliary tasks that preserve table's structural information by reconstructing the structure from GPT-2's representation and improving the text's fidelity with content matching task aligning the table and information in the generated text.By experimenting on Humans, Songs and Books, three few-shot table-to-text datasets in different domains, our model outperforms existing systems on most few-shot settings.

Research into the area of multiparty dialog has grown considerably over recent years.We present the Molweni dataset 1 , a machine reading comprehension (MRC) dataset with discourse structure built over multiparty dialog.Molweni's source samples from the Ubuntu Chat Corpus, including 10,000 dialogs comprising 88,303 utterances.We annotate 30,066 questions on this corpus, including both answerable and unanswerable questions.Molweni also uniquely contributes discourse dependency annotations in a modified Segmented Discourse Representation Theory (SDRT; (Asher et al., 2016)) style for all of its multiparty dialogs, contributing large-scale (78,245 annotated discourse relations) data to bear on the task of multiparty dialog discourse parsing.Our experiments show that Molweni is a challenging dataset for current MRC models: BERT-wwm, a current, strong SQuAD 2.0 performer, achieves only 67.7% F 1 on Molweni's questions, a 20+% significant drop as compared against its SQuAD 2.0 performance.

Emotion recognition in conversations (ERC) has received much attention recently in the natural language processing community.Considering that the emotions of the utterances in conversations are interactive, previous works usually implicitly model the emotion interaction between utterances by modeling dialogue context, but the misleading emotion information from context often interferes with the emotion interaction.We noticed that the gold emotion labels of the context utterances can provide explicit and accurate emotion interaction, but it is impossible to input gold labels at inference time.To address this problem, we propose an iterative emotion interaction network, which uses iteratively predicted emotion labels instead of gold emotion labels to explicitly model the emotion interaction.This approach solves the above problem, and can effectively retain the performance advantages of explicit modeling.We conduct experiments on two datasets, and our approach achieves state-of-the-art performance.

Deep contextualized word embeddings (Embeddings from Language Model, short for ELMo), as an emerging and effective replacement for the static word embeddings, have achieved success on a bunch of syntactic and semantic NLP problems. However, little is known about what is responsible for the improvements. In this article, we focus on the effect of ELMo for a typical syntax problem—universal POS tagging and dependency parsing. We incorporate ELMo as additional word embeddings into the state-of-the-art POS tagger and dependency parser, and it leads to consistent performance improvements. Experimental results show the model using ELMo outperforms the state-of-the-art baseline by an average of 0.91 for POS tagging and 1.11 for dependency parsing. Further analysis reveals that the improvements mainly result from the ELMo’s better abstraction ability on the out-of-vocabulary (OOV) words, and the character-level word representation in ELMo contributes a lot to the abstraction. Based on ELMo’s advantage on OOV, experiments that simulate low-resource settings are conducted and the results show that deep contextualized word embeddings are effective for data-insufficient tasks where the OOV problem is severe.

Many problems in natural language processing (NLP) can be cast as the problem of segmenting a sequence. In this article, we combine the semi-Markov conditional random fields (semi-CRF) with neural networks to solve NLP segmentation problems. We focus on the segment representation in neural semi-CRF which is important to the performance. Based on our preliminary work in Liu et al.[1], we represent a segment by both encoding the subsequence and embedding the segment string. We conduct a systematic study of the utility of various components in subsequence encoding and propose a method of constructing and deriving segment string embeddings. Extensive experiments on three typical segmentation problems, namely, shallow syntax parsing, named entity recognition, and Chinese word segmentation are conducted. The results show that we can achieve equally-performed subsequence encoding with a three times faster concatenation network compared to previous work. The results also show that the segment string embeddings help our neural semi-CRF model to achieve a macro-averaged error reduction of 13.15% over a strong baseline using deep contextualized embeddings and bidirectional long-short-term memory CRF, which also show the usefulness of semi-CRF even with contextualized embeddings. These results are competitive with the state-of-the-art segmentation systems.

Question answering (QA) and question generation (QG) are closely related tasks that could improve each other; however, the connection of these two tasks is not well explored in the literature. In this paper, we present two training algorithms for learning better QA and QG models through leveraging one another. The first algorithm extends Generative Adversarial Network (GAN), which selectively incorporates artificially generated instances as additional QA training data. The second algorithm is an extension of dual learning, which incorporates the probabilistic correlation of QA and QG as additional regularization in training objectives. To test the scalability of our algorithms, we conduct experiments on both document based and table based question answering tasks. Results show that both algorithms improve a QA model in terms of accuracy and QG model in terms of BLEU score. Moreover, we find that the performance of a QG model could be easily improved by a QA model via policy gradient, however, directly applying GAN that regards all the generated questions as negative instances could not improve the accuracy of the QA model. Our algorithm that selectively assigns labels to generated questions would bring a performance boost.

Detecting stance from certain types of question-answer pairs is an interesting problem which has not been carefully explored. Unlike previous stance detection tasks, targets here are not given entities or claims but entire questions, which makes it difficult to capture the semantics of targets and build target-dependent representations of answers. To address them, we introduce the Recurrent Conditional Attention (RCA) model which incorporates a conditional attention structure into the recurrent reading process. RCA iteratively guides the distillation of question semantic with answer information and collects stance-oriented text relating to question, further revealing mutual relationship among stance, answer and question. Experiments on a manually labeled Chinese community QA stance dataset show that RCA outperforms four strong baselines by average 2.90% on macro-F1 and 2.66% on micro-F1 respectively.

Machine reading comprehension with unanswerable questions is a challenging task.In this work, we propose a data augmentation technique by automatically generating relevant unanswerable questions according to an answerable question paired with its corresponding paragraph that contains the answer.We introduce a pair-to-sequence model for unanswerable question generation, which effectively captures the interactions between the question and the paragraph.We also present a way to construct training data for our question generation models by leveraging the existing reading comprehension dataset.Experimental results show that the pair-to-sequence model performs consistently better compared with the sequence-to-sequence baseline.We further use the automatically generated unanswerable questions as a means of data augmentation on the SQuAD 2.0 dataset, yielding 1.9 absolute F1 improvement with BERT-base model and 1.7 absolute F1 improvement with BERT-large model.

Scientists are relying heavily on biomedical literature search (BLS) engines (e.g., PubMed) to acquire knowledge. Existing BLS systems adopt a “C-A” paradigm that is to design query-document similarity measurement based on words/phrases in the unstructured Content and to develop search Algorithms. In this work, we argue that structures should be extracted and utilized to bridge the gap between text content and knowledge-based search. And graph is one of the most effective structured forms of knowledge and more informative than words or phrases. So we carry out a paradigm shift from “C-A” to “CTGA”. Here “T” is for factual tuple of concepts and relations, and “G” is for knowledge graph. Our proposed graph-based BLS system has three parts: (1) it uses neural information extraction models to turn text into tuples; (2) it represents the tuples of a query or a document as a knowledge graph of linked concept nodes; (3) it has an efficient graph-based matching algorithm to return related documents at the level of structured knowledge. Experiments show that in both objective and subjective evaluation, our CTGA performs significantly better than traditional CA-based PubMed.

Shuang Chen, Jinpeng Wang, Xiaocheng Feng, Feng Jiang, Bing Qin, Chin-Yew Lin. Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP). 2019.

Yiming Cui, Wanxiang Che, Ting Liu, Bing Qin, Shijin Wang, Guoping Hu. Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP). 2019.

Heng Gong, Xiaocheng Feng, Bing Qin, Ting Liu. Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP). 2019.

Tianwen Jiang, Tong Zhao, Bing Qin, Ting Liu, Nitesh Chawla, Meng Jiang. Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP). 2019.

Conditions play an essential role in scientific observations, hypotheses, and statements. Unfortunately, existing scientific knowledge graphs (SciKGs) represent factual knowledge as a flat relational network of concepts, as same as the KGs in general domain, without considering the conditions of the facts being valid, which loses important contexts for inference and exploration. In this work, we propose a novel representation of SciKG, which has three layers. The first layer has concept nodes, attribute nodes, as well as the attaching links from attribute to concept. The second layer represents both fact tuples and condition tuples. Each tuple is a node of the relation name, connecting to the subject and object that are concept or attribute nodes in the first layer. The third layer has nodes of statement sentences traceable to the original paper and authors. Each statement node connects to a set of fact tuples and/or condition tuples in the second layer. We design a semi-supervised Multi-Input Multi-Output sequence labeling model that learns complex dependencies between the sequence tags from multiple signals and generates output sequences for fact and condition tuples. It has a self-training module of multiple strategies to leverage the massive scientific data for better performance when manual annotation is limited. Experiments on a data set of 141M sentences show that our model outperforms existing methods and the SciKGs we constructed provide a good understanding of the scientific statements.

In recent years, recommendation systems have attracted more and more attention due to the rapid development of e-commerce. Reviews information can offer help in modeling user's preference and item's performance. Some existing methods utilize reviews for the recommendation. However, few of those models consider the importance of reviews and words in corpus together. Therefore, we propose an approach for rating prediction using a hierarchical attention-based network named HANN, which can distinguish the importance of reviews at both word level and review level for explanations automatically. Experiments on four real-life datasets from Amazon demonstrate that our model achieves an improvement in prediction compared to several state-of-the-art approaches. The hierarchical attention weights in sampled test data verify the effect on selecting informative words and reviews.

Facing lots of name mentions appearing on the web, entity linking is essential for many information processing applications. To improve linking accuracy, the relations between entities are usually considered in the linking process. This kind of method is called collective entity linking and can obtain high-quality results. There are two kinds of information helpful to reveal the relations between entities, i.e., contextual information and structural information of entities. Most traditional collective entity linking methods consider them separately. In fact, these two kinds of information represent entities from specific and diverse views and can enhance each other, respectively. Besides, if we look into each view closely, it can be separated into sub-views that are more meaningful. For this reason, this article proposes a multi-view–based collective entity linking algorithm, which combines several views of entities into an objective function for entity linking. The importance of each view can be valued and the linking results can be obtained along with resolving this objective function. Experimental results demonstrate that our linking algorithm can acquire higher accuracy than many state-of-the-art entity linking methods. Besides, since we simplify the entity's structure and change the entity linking to a sub-matrix searching problem, our algorithm also obtains high efficiency.

In this paper, we focus on multiple-choice reading comprehension which aims to answer a question given a passage and multiple candidate options. We present the hierarchical attention flow to adequately leverage candidate options to model the interactions among passages, questions and candidate options. We observe that leveraging candidate options to boost evidence gathering from the passages play a vital role in this task, which is ignored in previous works. In addition, we explicitly model the option correlations with attention mechanism to obtain better option representations, which are further fed into a bilinear layer to obtain the ranking score for each option. On a large-scale multiple-choice reading comprehension dataset (i.e. the RACE dataset), the proposed model outperforms two previous neural network baselines on both RACE-M and RACE-H subsets and yields the state-of-the-art overall results.

Many natural language processing tasks can be modeled into structured prediction and solved as a search problem.In this paper, we distill an ensemble of multiple models trained with different initialization into a single model.In addition to learning to match the ensemble's probability output on the reference states, we also use the ensemble to explore the search space and learn from the encountered states in the exploration.Experimental results on two typical search-based structured prediction tasks -transition-based dependency parsing and neural machine translation show that distillation can effectively improve the single model's performance and the final model achieves improvements of 1.32 in LAS and 2.65 in BLEU score on these two tasks respectively over strong baselines and it outperforms the greedy structured prediction models in previous literatures.

Yibo Sun, Duyu Tang, Nan Duan, Jianshu Ji, Guihong Cao, Xiaocheng Feng, Bing Qin, Ting Liu, Ming Zhou. Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2018.

Although end-to-end neural machine translation (NMT) has achieved remarkable progress in the recent years, the idea of adopting multipass decoding mechanism into conventional NMT is not well explored.In this paper, we propose a novel architecture called adaptive multi-pass decoder, which introduces a flexible multi-pass polishing mechanism to extend the capacity of NMT via reinforcement learning.More specifically, we adopt an extra policy network to automatically choose a suitable and effective number of decoding passes, according to the complexity of source sentences and the quality of the generated translations.Extensive experiments on Chinese-English translation demonstrate the effectiveness of our proposed adaptive multi-pass decoder upon the conventional NMT with a significant improvement about 1.55 BLEU.

In this paper, we propose a new rich resource enhanced AMR aligner which produces multiple alignments and a new transition system for AMR parsing along with its oracle parser.Our aligner is further tuned by our oracle parser via picking the alignment that leads to the highestscored achievable AMR graph.Experimental results show that our aligner outperforms the rule-based aligner in previous work by achieving higher alignment F1 score and consistently improving two open-sourced AMR parsers.Based on our aligner and transition system, we develop a transition-based AMR parser that parses a sentence into its AMR graph directly.An ensemble of our parsers with only words and POS tags as input leads to 68.4 Smatch F1 score, which outperforms the parser of Wang and Xue (2017).

Neural networks have been widely used for high resource language (e.g. English) named entity recognition (NER) and have shown state-of-the-art results.However, for low resource languages, such as Dutch, Spanish, due to the limitation of resources and lack of annotated data, taggers tend to have lower performances.To narrow this gap, we propose three novel strategies to enrich the semantic representations of low resource languages: we first develop neural networks to improve low resource word representations by knowledge transfer from high resource language using bilingual lexicons. Further, a lexicon extension strategy is designed to address out-of lexicon problem by automatically learning semantic projections.Thirdly, we regard word-level entity type distribution features as an external language-independent knowledge and incorporate them into our neural architecture. Experiments on two low resource languages (including Dutch and Spanish) demonstrate the effectiveness of these additional semantic representations (average 4.8\% improvement). Moreover, on Chinese OntoNotes 4.0 dataset, our approach achieved an F-score of 83.07\% with 2.91\% absolute gain compared to the state-of-the-art results.

We focus on essay generation, which is a challenging task that generates a paragraph-level text with multiple topics.Progress towards understanding different topics and expressing diversity in this task requires more powerful generators and richer training and evaluation resources. To address this, we develop a multi-topic aware long short-term memory (MTA-LSTM) network.In this model, we maintain a novel multi-topic coverage vector, which learns the weight of each topic and is sequentially updated during the decoding process.Afterwards this vector is fed to an attention model to guide the generator.Moreover, we automatically construct two paragraph-level Chinese essay corpora, 305,000 essay paragraphs and 55,000 question-and-answer pairs.Empirical results show that our approach obtains much better BLEU score compared to various baselines.Furthermore, human judgment shows that MTA-LSTM has the ability to generate essays that are not only coherent but also closely related to the input topics.

Yijia Liu, Yi Zhu, Wanxiang Che, Bing Qin, Nathan Schneider, Noah A. Smith. Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long Papers). 2018.

Distant supervised relation extraction (RE) has been an effective way of finding novel relational facts from text without labeled training data. Typically it can be formalized as a multi-instance multi-label problem.In this paper, we introduce a novel neural approach for distant supervised (RE) with specific focus on attention mechanisms.Unlike the feature-based logistic regression model and compositional neural models such as CNN, our approach includes two major attention-based memory components, which is capable of explicitly capturing the importance of each context word for modeling the representation of the entity pair, as well as the intrinsic dependencies between relations.Such importance degree and dependency relationship are calculated with multiple computational layers, each of which is a neural attention model over an external memory. Experiment on real-world datasets shows that our approach performs significantly and consistently better than various baselines.

Online users have generated a large amount of health-related data on medical forums and search engines. However, exploiting these rich data for orienting patient online and assisting medical checkup offline is nontrivial due to the sparseness of existing symptom-disease links, which caused by the natural and chatty expressions of symptoms. In this paper, we propose a novel and general representation learning method ContextCare for human generated health-related data, which learns the latent relationship between symptoms and diseases from the symptom-disease diagnosis network for disease prediction, disease category prediction and disease clustering. To alleviate the network sparseness, ContextCare adopts regularizations from rich contextual information networks including a symptom co-occurrence network and a disease evolution network. Therefore, our representations of symptoms and diseases incorporate knowledge from these three networks. Extensive experiments on medical forum data demonstrate that ContextCare outperforms the state-of-the-art methods in disease category prediction, disease prediction and disease clustering.

In this paper, we formally define the problem of representing and leveraging abstract event causality to power downstream applications. We propose a novel solution to this problem, which build an abstract causality network and embed the causality network into a continuous vector space. The abstract causality network is generalized from a specific one, with abstract event nodes represented by frequently co-occurring word pairs. To perform the embedding task, we design a dual cause-effect transition model. Therefore, the proposed method can obtain general, frequent, and simple causality patterns, meanwhile, simplify event matching. Given the causality network and the learned embeddings, our model can be applied to a wide range of applications such as event prediction, event clustering and stock market movement prediction. Experimental results demonstrate that 1) the abstract causality network is effective for discovering high-level causality rules behind specific causal events; 2) the embedding models perform better than state-of-the-art link prediction techniques in predicting events; and 3) the event causality embedding is an easy-to-use and sophisticated feature for downstream applications such as stock market movement prediction.

Knowledge base (KB) such as Freebase plays an important role for many natural language processing tasks. English knowledge base is obviously larger and of higher quality than low resource language like Chinese. To expand Chinese KB by leveraging English KB resources, an effective way is to translate English KB (source) into Chinese (target). In this direction, two major challenges are to model triple semantics and to build a robust KB translator. We address these challenges by presenting a neural network approach, which learns continuous triple representation with a gated neural network. Accordingly, source triples and target triples are mapped in the same semantic vector space. We build a new dataset for English-Chinese KB translation from Freebase, and compare with several baselines on it. Experimental results show that the proposed method improves translation accuracy compared with baseline methods. We show that adaptive composition model improves standard solution such as neural tensor network in terms of translation accuracy.

Target-dependent sentiment classification remains a challenge: modeling the semantic relatedness of a target with its context words in a sentence. Different context words have different influences on determining the sentiment polarity of a sentence towards the target. Therefore, it is desirable to integrate the connections between target word and context words when building a learning system. In this paper, we develop two target dependent long short-term memory (LSTM) models, where target information is automatically taken into account. We evaluate our methods on a benchmark dataset from Twitter. Empirical results show that modeling sentence representation with standard LSTM does not perform well. Incorporating target information into LSTM can significantly boost the classification accuracy. The target-dependent LSTM models achieve state-of-the-art performances without using syntactic parser or external sentiment lexicons.

We introduce a deep memory network for aspect level sentiment classification.Unlike feature-based SVM and sequential neural models such as LSTM, this approach explicitly captures the importance of each context word when inferring the sentiment polarity of an aspect.Such importance degree and text representation are calculated with multiple computational layers, each of which is a neural attention model over an external memory.Experiments on laptop and restaurant datasets demonstrate that our approach performs comparable to state-of-art feature based SVM system, and substantially better than LSTM and attention-based LSTM architectures.On both datasets we show that multiple computational layers could improve the performance.Moreover, our approach is also fast.The deep memory network with 9 layers is 15 times faster than LSTM with a CPU implementation.

We propose learning sentiment-specific word embeddings dubbed sentiment embeddings in this paper. Existing word embedding learning algorithms typically only use the contexts of words but ignore the sentiment of texts. It is problematic for sentiment analysis because the words with similar contexts but opposite sentiment polarity, such asgoodandbad, are mapped to neighboring word vectors. We address this issue by encoding sentiment information of texts (e.g., sentences and words) together with contexts of words in sentiment embeddings. By combining context and sentiment level evidences, the nearest neighbors in sentiment embedding space are semantically similar and it favors words with the same sentiment polarity. In order to learn sentiment embeddings effectively, we develop a number of neural networks with tailoring loss functions, and collect massive texts automatically with sentiment signals like emoticons as the training data. Sentiment embeddings can be naturally used as word features for a variety of sentiment analysis tasks without feature engineering. We apply sentiment embeddings to word-level sentiment analysis, sentence level sentiment classification, and building sentiment lexicons. Experimental results show that sentiment embeddings consistently outperform context-based embeddings on several benchmark datasets of these tasks. This work provides insights on the design of neural networks for learning task-specific word embeddings in other natural language processing tasks.

Duyu Tang, Bing Qin, Ting Liu. Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Processing (Volume 1: Long Papers). 2015.

Muyu Zhang, Bing Qin, Mao Zheng, Graeme Hirst, Ting Liu. Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Processing (Volume 1: Long Papers). 2015.

Document level sentiment classification remains a challenge: encoding the intrin-sic relations between sentences in the se-mantic meaning of a document. To ad-dress this, we introduce a neural network model to learn vector-based document rep-resentation in a unified, bottom-up fash-ion. The model first learns sentence rep-resentation with convolutional neural net-work or long short-term memory. After-wards, semantics of sentences and their relations are adaptively encoded in docu-ment representation with gated recurren-t neural network. We conduct documen-t level sentiment classification on four large-scale review datasets from IMDB and Yelp Dataset Challenge. Experimen-tal results show that: (1) our neural mod-el shows superior performances over sev-eral state-of-the-art algorithms; (2) gat-ed recurrent neural network dramatically outperforms standard recurrent neural net-work in document modeling for sentiment classification.1 1

Syntactic linearization algorithms take a bag of input words and a set of optional constraints, and construct an output sentence and its syntactic derivation simultaneously.The search problem is NP-hard, and the current best results are achieved by bottom-up bestfirst search.One drawback of the method is low efficiency; and there is no theoretical guarantee that a full sentence can be found within bounded time.We propose an alternative algorithm that constructs output structures from left to right using beam-search.The algorithm is based on incremental parsing algorithms.We extend the transition system so that word ordering is performed in addition to syntactic parsing, resulting in a linearization system that runs in guaranteed quadratic time.In standard evaluations, our system runs an order of magnitude faster than a state-of-the-art baseline using best-first search, with improved accuracies.

Muyu Zhang, Vanessa Wei Feng, Bing Qin, Graeme Hirst, Ting Liu, Jingwen Huang. Proceedings of the 2015 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. 2015.

Current online public opinion analysis systems can explore lots of hot events and present the public emotion distribution for each event, which are useful for the governments and companies. However, the public emotion distributions are just the shallow analysis of the hot events, more and more people want to know the hidden causation behind the emotion distributions. Thus, this paper presents a deep Event-Emotion analysis system on Microblogs to reveal what causes different emotions of a hot event. We here use several related sub-events to describe a hot event in different perspectives, accordingly these sub-events combined with their different emotion distributions can be used to explain the total emotion distribution of a hot event. Experiments on 15 hot events show that the above idea is reasonable to exploit the emotion causation and can help people better understand the evolution of the hot event. Furthermore, this deep Event-Emotion analysis system also tracks the amount treads and emotion treads of the hot event, and presents the deep analysis based on the user profile.

Traditional event extraction systems focus mainly on event type identification and event participant extraction based on pre-specified event type paradigms and manually annotated corpora. However, different domains have different event type paradigms. When transferring to a new domain, we have to build a new event type paradigm and annotate a new corpus from scratch. This kind of conventional event extraction system requires massive human effort, and hence prevents event extraction from being widely applicable. In this paper, we present BUEES, a bottom-up event extraction system, which extracts events from the web in a completely unsupervised way. The system automatically builds an event type paradigm in the input corpus, and then proceeds to extract a large number of instance patterns of these events. Subsequently, the system extracts event arguments according to these patterns. By conducting a series of experiments, we demonstrate the good performance of BUEES and compare it to a state-of-the-art Chinese event extraction system, i.e., a supervised event extraction system. Experimental results show that BUEES performs comparably (5% higher F-measure in event type identification and 3% higher F-measure in event argument extraction), but without any human effort.

Semantic hierarchy construction aims to build structures of concepts linked by hypernym-hyponym (“is-a”) relations. A major challenge for this task is the automatic discovery of such relations. This paper proposes a novel and effective method for the construction of semantic hierarchies based on continuous vector representation of words, named word embeddings, which can be used to measure the semantic relationship between words. We identify whether a candidate word pair has hypernym-hyponym relation by using the word-embedding-based semantic projections between words and their hypernyms. Our result, an F-score of 73.74%, outperforms the state-of-the-art methods on a manually labeled test dataset. Moreover, combining our method with a previous manually built hierarchy extension method can further improve F-score to 80.29%.

In this paper, we propose a joint segmentation and classification framework for sentence-level sentiment classification. It is widely recognized that phrasal information is crucial for sentiment classification. However, existing sentiment classification algorithms typically split a sentence as a word sequence, which does not effectively handle the inconsistent sentiment polarity between a phrase and the words it contains, such as {“not bad,” “bad”} and {“a great deal of,” “great”}. We address this issue by developing a joint framework for sentence-level sentiment classification. It simultaneously generates useful segmentations and predicts sentence-level polarity based on the segmentation results. Specifically, we develop a candidate generation model to produce segmentation candidates of a sentence; a segmentation ranking model to score the usefulness of a segmentation candidate for sentiment classification; and a classification model for predicting the sentiment polarity of a segmentation. We train the joint framework directly from sentences annotated with only sentiment polarity, without using any syntactic or sentiment annotations in segmentation level. We conduct experiments for sentiment classification on two benchmark datasets: a tweet dataset and a review dataset. Experimental results show that: 1) our method performs comparably with state-of-the-art methods on both datasets; 2) joint modeling segmentation and classification outperforms pipelined baseline methods in various experimental settings.

Semantic hierarchy construction aims to build structures of concepts linked by hypernym-hyponym ("is-a") relations.A major challenge for this task is the automatic discovery of such relations.This paper proposes a novel and effective method for the construction of semantic hierarchies based on word embeddings, which can be used to measure the semantic relationship between words.We identify whether a candidate word pair has hypernym-hyponym relation by using the word-embedding-based semantic projections between words and their hypernyms.Our result, an F-score of 73.74%, outperforms the state-of-theart methods on a manually labeled test dataset.Moreover, combining our method with a previous manually-built hierarchy extension method can further improve Fscore to 80.29%.

We present a method that learns word embedding for Twitter sentiment classification in this paper.Most existing algorithms for learning continuous word representations typically only model the syntactic context of words but ignore the sentiment of text.This is problematic for sentiment analysis as they usually map words with similar syntactic context but opposite sentiment polarity, such as good and bad, to neighboring word vectors.We address this issue by learning sentimentspecific word embedding (SSWE), which encodes sentiment information in the continuous representation of words.Specifically, we develop three neural networks to effectively incorporate the supervision from sentiment polarity of text (e.g.sentences or tweets) in their loss functions.To obtain large scale training corpora, we learn the sentiment-specific word embedding from massive distant-supervised tweets collected by positive and negative emoticons.Experiments on applying SS-WE to a benchmark Twitter sentiment classification dataset in SemEval 2013 show that (1) the SSWE feature performs comparably with hand-crafted features in the top-performed system; (2) the performance is further improved by concatenating SSWE with existing feature set.

In this paper, we propose a joint segmentation and classification framework for sentiment analysis.Existing sentiment classification algorithms typically split a sentence as a word sequence, which does not effectively handle the inconsistent sentiment polarity between a phrase and the words it contains, such as "not bad" and "a great deal of ".We address this issue by developing a joint segmentation and classification framework (JSC), which simultaneously conducts sentence segmentation and sentence-level sentiment classification.Specifically, we use a log-linear model to score each segmentation candidate, and exploit the phrasal information of top-ranked segmentations as features to build the sentiment classifier.A marginal log-likelihood objective function is devised for the segmentation model, which is optimized for enhancing the sentiment classification performance.The joint model is trained only based on the annotated sentiment polarity of sentences, without any segmentation annotations.Experiments on a benchmark Twitter sentiment classification dataset in SemEval 2013 show that, our joint model performs comparably with the state-of-the-art methods.

Web users are increasingly looking for structured data, such as lyrics, job, or recipes, using unstructured queries on the web.However, retrieving relevant results from such data is a challenging problem due to the unstructured language of the web queries.In this paper, we propose a method to improve web search ranking by detecting Structured Annotation of queries based on top search results.In a structured annotation, the original query is split into different units that are associated with semantic attributes in the corresponding domain.We evaluate our techniques using real world queries and achieve significant improvement.

Hypernym discovery aims to extract such noun pairs that one noun is a hypernym of the other.Most previous methods are based on lexical patterns but perform badly on opendomain data.Other work extracts hypernym relations from encyclopedias but has limited coverage.This paper proposes a simple yet effective distant supervision framework for Chinese open-domain hypernym discovery.Given an entity name, we try to discover its hypernyms by leveraging knowledge from multiple sources, i.e., search engine results, encyclopedias, and morphology of the entity name.First, we extract candidate hypernyms from the above sources.Then, we apply a statistical ranking model to select correct hypernyms.A set of novel features is proposed for the ranking model.We also present a heuristic strategy to build a large-scale noisy training data for the model without human annotation.Experimental results demonstrate that our approach outperforms the state-of-the-art methods on a manually labeled test dataset.

Linking name mentions in microblog posts to a knowledge base, namely microblog entity linking, is useful for text mining tasks on microblog.Entity linking in long text has been well studied in previous works.However few work has focused on short text such as microblog post.Microblog posts are short and noisy.Previous method can extract few features from the post context.In this paper we propose to use extra posts for the microblog entity linking task.Experimental results show that our proposed method significantly improves the linking accuracy over traditional methods by 8.3% and 7.5% respectively.