Linrui Ma, Chun Hei Lo, Xinyu Wang, Peng Lu, Xihao Yuan, Hanting Chen, Kai Han, Xinghao Chen, Chengjun Zhan, Hanlin xu, Yichun Yin, Lifeng Shang, Feng Wen, Boxing Chen, Yufei Cui. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Xueqing Peng, Lingfei Qian, Yan Wang, Ruoyu Xiang, Yueru He, Yang Ren, Mingyang Jiang, Vincent Jim Zhang, Yuqing Guo, Jeff Zhao, Huan He, Yi Han, Yun Feng, Yuechen Jiang, Yupeng Cao, Haohang Li, Yangyang Yu, Xiaoyu Wang, Penglei Gao, Shengyuan Lin, Keyi Wang, Shanshan Yang, Yilun Zhao, Zhiwei Liu, Peng Lu, Jerry Huang, Suyuchen Wang, Triantafillos Papadopoulos, Polydoros Giannouris, Efstathia Soufleri, Nuo Chen, Zhiyang Deng, Heming Fu, Yijia Zhao, Mingquan Lin, Meikang Qiu, Kaleb E Smith, Arman Cohan, Xiao-Yang Liu, Jimin Huang, Guojun Xiong, Alejandro Lopez-Lira, Xi Chen, Junichi Tsujii, Jian-Yun Nie, Sophia Ananiadou, Qianqian Xie. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Retrieval-Augmented Generation (RAG) is becoming increasingly essential for Question Answering (QA) in the financial sector, where accurate and contextually grounded insights from complex public disclosures are crucial. However, existing financial RAG systems face two significant challenges: (1) they struggle to process heterogeneous data formats, such as text, tables, and figures; and (2) they encounter difficulties in balancing general-domain applicability with company-specific adaptation. To overcome these challenges, we present VeritasFi, an innovative hybrid RAG framework that incorporates a multi-modal preprocessing pipeline alongside a cutting-edge two-stage training strategy for its re-ranking component. VeritasFi enhances financial QA through three key innovations: (1) A multi-modal preprocessing pipeline that seamlessly transforms heterogeneous data into a coherent, machine-readable format. (2) A tripartite hybrid retrieval engine that operates in parallel, combining deep multi-path retrieval over a semantically indexed document corpus, real-time data acquisition through tool utilization, and an expert-curated memory bank for high-frequency questions, ensuring comprehensive scope, accuracy, and efficiency. (3) A two-stage training strategy for the document re-ranker, which initially constructs a general, domain-specific model using anonymized data, followed by rapid fine-tuning on company-specific data for targeted applications. By integrating our proposed designs, VeritasFi presents a novel framework that greatly enhances the adaptability and robustness of financial RAG systems, providing a scalable solution for both general-domain and company-specific QA tasks. Code accompanying this work is available at https://github.com/simplew4y/VeritasFi.git.

Aspect-based sentiment analysis (ABSA) aims to extract fine-grained opinions from the text by discerning sentiments toward specific aspects. Although large language models (LLMs) perform well in-context learning (ICL), current ICL methodologies typically retrieve semantically similar but structurally redundant examples, failing to capture syntactic and aspect-level cues critical for ABSA. To overcome these limitations, we report Multi-perspective Sequential retrieval with Predictive Feedback (MSPF), a few-shot learning framework that enhances ICL through MSPF, which integrates three complementary perspectives: overall semantic, syntactic relevance, and aspect sentiment alignment. Evaluated on four benchmark datasets (Laptop14, Restaurant14, Books, and Clothing), MSPF achieved F1 scores of 67.03 % (Laptop14), 73.51 % (Restaurant14), 76.07 % (Books), and 81.96 % (Clothing), outperforming standard ICL by +7.06 %, +5.60 %, +25.61 %, and +18.38 %, respectively. These results validated the efficacy of MSPF in improving LLM reasoning for fine-grained sentiment tasks with limited annotations.

Leveraging large language models in real-world settings often entails a need to utilize domain-specific data and tools in order to follow the complex regulations that need to be followed for acceptable use. Within financial sectors, modern enterprises increasingly rely on Retrieval-Augmented Generation (RAG) systems to address complex information retrieval in financial document workflows. However, existing solutions struggle to account for the inherent heterogeneity of data (e.g., text, tables, diagrams) and evolving complexity in financial filings, leading to compromised accuracy in critical information extraction. We propose the FinSage framework as a solution, utilizing a multi-aspect RAG framework tailored for data retrieval and summarization in multi-modal financial documents. øurmodel introduces three innovative components: (1) a multi-modal pre-processing pipeline that unifies diverse data formats and generates chunk-level metadata summaries, (2) a multi-path sparse-dense retrieval system augmented with query expansion (HyDE) and metadata-aware semantic search, and (3) a domain-specialized re-ranking module fine-tuned via Direct Preference Optimization to prioritize ground-truth-related content. Extensive experiments demonstrate that FinSage achieves an impressive recall of 92.51% on 75 expert-curated questions derived from surpasses the best baseline method on the FinanceBench question answering datasets by 24.06% in accuracy. Moreover, FinSage has been successfully deployed as financial question-answering system in online meetings, where it has already served more than 1,200 people. The implementation is publicly available at https://github.com/simplew4y/finsage.

Multi-task learning has demonstrated remarkable success across a broad spectrum of natural language processing tasks, particularly with neural network-based methods. Despite these advances, a fundamental gap remains in explaining the relationship between task-relatedness and model effectiveness. To address this issue, we propose a novel approach for implicitly modeling task-relatedness by leveraging a quantum physical mathematical framework. In this paper, we introduce a complex-valued neural network designed to encapsulate and analyze task-relatedness. Within this framework, sentences originating from diverse tasks are encoded as mixed quantum systems, represented on a meticulously defined Semantic Hilbert Space. This allows the network to interpret inter-task relationships through the explicit physical semantics of well-constrained components grounded in quantum probability theory. By adhering to these rigorous principles, our model not only establishes a robust method for quantifying task-relatedness but also fosters a deeper, self-explanatory understanding of the underlying processes. To validate the efficacy of our approach, we conducted extensive experiments across five benchmark text classification tasks. The results demonstrate both the superior performance and the interpretability of the proposed model, highlighting its potential as a self-explanatory system for multi-task learning in NLP.

Large Language Models (LLMs) have demonstrated remarkable performance across various natural language processing (NLP) tasks. However, their deployment is challenging due to the substantial computational resources required. Power-of-two (PoT) quantization is a general tool to counteract this difficulty. Albeit previous works on PoT quantization can be efficiently dequantized on CPUs using fixed-point addition, it showed less effectiveness on GPUs. The reason is entanglement of the sign bit and sequential bit manipulations needed for dequantization. We propose a novel POT quantization framework for LLM weights that (i) outperforms state-of-the-art accuracy in extremely low-precision number formats, and (ii) enables faster inference through more efficient dequantization. To maintain the accuracy of the quantized model, we introduce a two-step post-training algorithm: (i) initialize the quantization scales with a robust starting point, and (ii) refine these scales using a minimal calibration set. The performance of our PoT post-training algorithm surpasses the current state-of-the-art in integer quantization, particularly at low precisions such as 2- and 3-bit formats. Our PoT quantization accelerates the dequantization step required for the floating point inference and leads to 3.67× speed up on a NVIDIA V100, and 1.63× on a NVIDIA RTX 4090, compared to uniform integer dequantization.

Over recent years, the Transformer has become a fundamental building block for sequence modeling architectures. Yet at its core is the use of self-attention, whose memory and computational cost grow quadratically with the sequence length $N$, rendering it prohibitively expensive for long sequences. A promising approach is top-$k$ attention, which selects only the $k$ most relevant tokens and achieves performance comparable to vanilla self-attention while significantly reducing space and computational demands. However, causal masks require the current query token to only attend to past tokens, preventing existing top-$k$ attention methods from efficiently searching for the most relevant tokens in parallel, thereby limiting training efficiency. In this work, we propose ZETA, leveraging Z-Order Curves for Efficient Top-k Attention, to enable parallel querying of past tokens for entire sequences. We first theoretically show that the choice of key and query dimensions involves a trade-off between the curse of dimensionality and the preservation of relative distances after projection. In light of this insight, we propose reducing the dimensionality of keys and queries in contrast to values and further leveraging Z-order curves to map low-dimensional keys and queries into one-dimensional space, which permits parallel sorting, thereby largely improving the efficiency for top-$k$ token selection. Experimental results demonstrate that ZETA~matches the performance of standard attention on synthetic tasks Associative Recall and outperforms attention and its variants on Long-Range Arena and WikiText-103 language modeling.

Recent advances in natural language processing (NLP) have opened up greater opportunities to enable fine-tuned large language models (LLMs) to behave as more powerful interactive agents through improved instruction-following ability. However, understanding how this impacts confidence calibration for reliable model output has not been researched in full. In this work, we examine various open-sourced LLMs, identifying significant calibration degradation after instruction tuning in each. Seeking a practical solution, we look towards label smoothing, which has been shown as an effective method to regularize for overconfident predictions but has yet to be widely adopted in the supervised fine-tuning (SFT) of LLMs. We first provide insight as to why label smoothing is sufficient to maintain calibration throughout the SFT process. However, settings remain where the effectiveness of smoothing is severely diminished, in particular the case of large vocabulary LLMs (LV-LLMs). We posit the cause to stem from the ability to become over-confident, which has a direct relationship with the hidden size and vocabulary size, and justify this theoretically and experimentally. Finally, we address an outstanding issue regarding the memory footprint of the cross-entropy loss computation in the label smoothed loss setting, designing a customized kernel to dramatically reduce memory consumption without sacrificing speed or performance in comparison to existing solutions for non-smoothed losses.

Peng Lu, Ivan Kobyzev, Mehdi Rezagholizadeh, Boxing Chen, Philippe Langlais. 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.

The quadratic complexity of the attention mechanism in Transformer models has motivated the development of alternative architectures with sub-quadratic scaling, such as state-space models. Among these, Mamba has emerged as a leading architecture, achieving state-of-the-art results across a range of language modeling tasks. However, Mamba’s performance significantly deteriorates when applied to contexts longer than those seen during pre-training, revealing a sharp sensitivity to context length extension. Through detailed analysis, we attribute this limitation to the out-of-distribution behavior of its state-space dynamics, particularly within the parameterization of the state transition matrix $A$. Unlike recent works which attribute this sensitivity to the vanished accumulation of discretization time steps, $\exp(-\sum_{t=1}^N{\Delta}_t)$, we establish a connection between state convergence behavior as the input length approaches infinity and the spectrum of the transition matrix $A$, offering a well-founded explanation of its role in length extension. Next, to overcome this challenge, we propose an approach that applies spectrum scaling to pre-trained Mamba models to enable robust long-context generalization by selectively modulating the spectrum of $A$ matrices in each layer. We show that this can significantly improve performance in settings where simply modulating ${\Delta}_t$ fails, validating our insights and providing avenues for better length generalization of state-space models with structured transition matrices.

Recent advancements in online social networks (e.g., Facebook, Twitter, Weibo) highlight the need for effective interventions in combating the rapid spread of negative emotions triggered by adverse public events. Existing studies on emotional contagion have explored factors like network structure and individual susceptibility but lack a systematic understanding of intervention measures and their efficacy. This article addresses this gap as the negative emotion intervention (NEI) problem. While current strategies, such as blocking users, struggle with accurately interpreting subtle shifts in contagion patterns, our study focuses on leveraging reinforcement learning with deep Q-networks (DQN) to develop dynamic blocking strategies. We introduce the novel negative emotion contagion model (NECM) that considers spreaders’ capability, receivers’ sensitivity, and event popularity. Our study pioneers the optimized dynamic blocking approach, NECM-RLDB, which outperforms heuristic methods (out-degree, betweenness centrality, PageRank) in experimental tests on artificial and real-world datasets.

Ivan Kobyzev, Aref Jafari, Mehdi Rezagholizadeh, Tianda Li, Alan Do-Omri, Peng Lu, Pascal Poupart, Ali Ghodsi. Proceedings of the 17th Conference of the European Chapter of the Association for Computational Linguistics. 2023.

Transformer-based models have achieved stateof-the-art performance on numerous NLP applications.However, long documents which are prevalent in real-world scenarios cannot be efficiently processed by transformers with the vanilla self-attention module due to their quadratic computation complexity and limited length extrapolation ability.Instead of tackling the computation difficulty for self-attention with sparse or hierarchical structures, in this paper, we investigate the use of State-Space Models (SSMs) for long document classification tasks.We conducted extensive experiments on six long document classification datasets, including binary, multi-class, and multi-label classification, comparing SSMs (with and without pre-training) to self-attention-based models.We also introduce the SSM-pooler model and demonstrate that it achieves comparable performance while being on average 36% more efficient.Additionally our method exhibits higher robustness to the input noise even in the extreme scenario of 40%. * Research done during internship in Huawei Noah's Ark Lab (Montreal).

In recent years, COVID‐19 has become the hottest topic. Various issues, such as epidemic transmission routes and preventive measures, have “occupied” several online social media platforms. Many rumors about COVID‐19 have also arisen, causing public anxiety and seriously affecting normal social order. Identifying a rumor at its very inception is crucial to reducing the potential harm of its evolution to society as a whole. However, epidemic rumors provide limited signal features in the early stage. In order to identify rumors with data sparsity, we propose a few‐shot learning rumor detection model based on capsule networks (CNFRD), utilizing the metric learning framework and the capsule network to detect the rumors posted during unexpected epidemic events. Specifically, we constructively use the capsule network neural layer to summarize the historical rumor data and obtain the generalized class representation based on the historical rumor data samples. Besides, we calculate the distance between the epidemic rumor sample and the historical rumor class‐wise representation according to the metric module. Finally, epidemic rumors are discriminated against according to the nearest neighbor principle. The experimental results prove that the proposed method can achieve higher accuracy with fewer epidemic rumor samples. This approach provided 88.92% accuracy on the Chinese rumor dataset and 87.07% accuracy on the English rumor dataset, which improved by 7% to 23% over existing approaches. Therefore, the CNFRD model can identify epidemic rumors in COVID‐19 as early as possible and effectively improve the performance of rumor detection.