Large Language Models (LLMs) have achieved remarkable success in recent years, owing to their impressive generalization capabilities and rich world knowledge. To capitalize on the potential of using LLMs as recommender systems, mainstream approaches typically focus on two paradigms. The first paradigm designs multi-domain or multi-task instruction data for generalizable recommendation, so as to align LLMs with general recommendation areas and deal with cold-start recommendation. The second paradigm focuses on enhancing domain-specific recommendation tasks, improving performance in warm recommendation scenarios. While most previous works treat these two paradigms separately, we argue that they have complementary advantages, and combining them can yield better results. In this paper, we propose a generalizable and efficient LLM-based recommendation framework RecCocktail. Our approach begins with fine-tuning a "base spirit" LoRA module using domain-general recommendation instruction data to align LLM with recommendation knowledge. Next, given users' behavior of a specific domain, we construct a domain-specific "ingredient" LoRA module. We then provide an entropy-guided adaptive merging method to mix the "base spirit" and the "ingredient" in the weight space. Please note that, RecCocktail combines the advantages of the existing two paradigms without introducing additional time or space overhead during the inference phase. Moreover, RecCocktail is efficient with plug and play, as the "base spirit" LoRA is trained only once, and any domain-specific "ingredient" can be efficiently mixed with only domain-specific fine-tuning. Extensive experiments on multiple datasets under both warm and cold-start recommendation scenarios validate the effectiveness and generality of the proposed RecCocktail.

Multimodal Document Question Answering (MDQA) systems commonly follow a retrieve-then-answer paradigm; however, the retrieval stage often introduces substantial noise, making an effective reranking component indispensable. Existing reranker training frameworks in MDQA rely predominantly on proxy supervision derived from human annotations or large language model (LLM) outputs, which are frequently noisy and, more critically, misaligned with downstream answer quality. To overcome this limitation, we propose AD-Reranker, a novel framework that shifts reranker training from proxy imitation to answer-driven utility optimization. Specifically, we reformulate the reranker as an environment-grounded agent that interacts with a downstream reader, modeled as a deterministic environment. We further design a composite reward function that integrates answer correctness, thereby explicitly incentivizing ranking strategies aligned with downstream task performance. To optimize the agent, we adopt Group Relative Policy Optimization (GRPO), enabling stable and effective group-wise policy learning. Empirical results demonstrate that AD-Reranker achieves superior reranking quality and an optimal accuracy-efficiency trade-off. When integrated into standard MDQA pipelines, AD-Reranker improves accuracy by 1.9%–5.0% while reducing the reader's context token consumption by 15%–52%, providing strong evidence for the effectiveness of answer-driven reranker training.

Implicit feedback is central to modern recommender systems but is inherently noisy, often impairing model training and degrading user experience. At scale, such noise can mislead learning processes, reducing both recommendation accuracy and platform value. Existing denoising strategies typically overlook the entity-specific nature of noise while introducing high computational costs and complex hyperparameter tuning. To address these challenges, we propose EARD (Entity-Aware Reliability-Driven Denoising), a lightweight framework that shifts the focus from interaction-level signals to entity-level reliability. Motivated by the empirical observation that training loss correlates with noise, EARD quantifies user and item reliability via their average training losses as a proxy for reputation, and integrates these entity-level factors with interaction-level confidence. The framework is model-agnostic, computationally efficient, and requires only two intuitive hyperparameters. Extensive experiments across multiple datasets and backbone models demonstrate that EARD yields substantial improvements over state-of-the-art baselines (e.g., up to 27.01% gain in NDCG@50), while incurring negligible additional computational cost. Comprehensive ablation studies and mechanism analyses further confirm EARD's robustness to hyperparameter choices and its practical scalability. These results highlight the importance of entity-aware reliability modeling for denoising implicit feedback and pave the way for more robust recommendation research.

In web-based AI services, providers typically host multiple large language models (LLMs) that exhibit diverse capabilities and incur different API costs. Meanwhile, LLM's performance depends not only on its inherent capacity but also on the reasoning strategy it employs, which together influence both answer quality and computational cost. A key challenge is therefore how to adaptively allocate models and strategies to achieve high-quality responses under constrained costs. To address this challenge, we propose Route-To-Reason (RTR), a unified routing framework that simultaneously selects suitable LLMs and reasoning strategies according to query complexity and user budget. Specifically, RTR learns dense vector representations of models and strategies that capture their behavioral characteristics in handling different queries. Leveraging these embeddings, RTR builds a routing table that estimates the cost and performance of different model–strategy pairs. During inference, RTR consults this routing table to dynamically assign the most appropriate pair, enabling adaptive and cost-efficient reasoning tailored to query difficulty and budget scenarios. Extensive experiments across multiple reasoning benchmarks show that RTR achieves comparable or higher accuracy than the best single LLM while substantially reducing both token usage and API cost (by up to 60%), achieving a superior trade-off between performance and efficiency. By lowering the overhead of large-scale LLM inference, RTR contributes to cost-aware and environmentally sustainable deployment of web-based AI services.

Large-scale online marketplaces and recommender systems are crucial technological foundations for the development of e-commerce. In industrial recommender systems, features play a vital role as they carry essential information for downstream models. Accurate estimation of feature importance is critical, as it helps identify the most useful feature subsets from thousands of candidates for online services. Such a selection enables optimization of online performance while reducing computational burden. To address the feature selection challenges in deep learning, trainable gate-based and sensitivity-based methods have been proposed and proven effective in the industry. However, by analyzing real-world examples, we identified three bias issues that cause feature importance estimation to rely on partial model layers, samples, or gradients, ultimately leading to inaccurate feature importance estimates. We refer to these biases as layer bias, baseline bias, and approximation bias. To mitigate these biases, we propose FairFS, a fair and accurate feature selection algorithm. On one hand, FairFS directly regularizes feature importance estimation across all non-linear transformational layers to avoid layer bias. On the other hand, it employs a smooth baseline feature close to the classifier's decision boundary and an aggregated approximation method to mitigate bias issues. Extensive experiments demonstrate how FairFS mitigates these three biases and achieves state-of-the-art feature selection results.

Science Question Answering (SQA) is an important task for evaluating models' capability to reason with scientific knowledge. However, the extensive availability of scientific information (e.g., basic concepts in biology, physics, and chemistry) in pre-trained corpora may cause large language models (LLMs) to rely more on memorized information rather than actual reasoning when answering questions. This reliance persists even with techniques like Chain-of-Thought prompting, resulting in shallow understanding and limited reasoning based on scientific knowledge. Therefore, to enhance LLMs' capacity to comprehend and apply scientific knowledge, we propose a framework calledMulti-AgentCooperation-basedKnowledgeExemplification (MCKE). Specifically, MCKE leverages knowledge alongside questions to create exemplified knowledge, promoting deeper understanding through innovative knowledge representation. To better evaluate the model's ability to reason and apply knowledge, we introduceNovSciQA, a multiple-choice question answering dataset based on newly created scientific knowledge. This dataset covers multi-subject scientific knowledge and questions that do not exist in reality, making it impossible for the model to rely on memorized answer-related information to answer questions. Experimental results show that the MCKE framework outperforms baselines, and the NovSciQA dataset effectively assesses models' knowledge understanding and application. Our code and dataset are available inhttps://anonymous.4open.science/r/MCKE-NovSciQA.

The recent advancements in Deep Reinforcement Learning (DRL) have significantly improved the performance of adaptive Traffic Signal Control (TSC). However, DRL policies are typically represented by over-parameterized neural networks, which function as black-box models. Consequently, the learned policies often lack interpretability and are challenging to deploy on resource-constrained edge hardware. Moreover, the DRL methods frequently exhibit poor generalization, struggling to transfer the learned policies across different geographical regions. These limitations hinder the real-world applicability of learning-based approaches. To address these issues, we suggest the use of an inherently interpretable program for representing the control policy. We present Programmatic Interpretable reinforcement learning for effective traffic signal control ($\pi$-eLight), a new approach designed to autonomously discover non-differentiable programs. Specifically, we first define an effective program framework as the control policy, where certain components remain learnable. Next, we introduce a Domain Specific Language (DSL) for constructing interpretable programs and transformation rules for generating programs with hierarchical structures. Last, we utilize Monte Carlo Tree Search (MCTS) to find the optimal program in a discrete space. Extensive experiments demonstrate that$\pi$-eLight consistently outperforms DRL-based baselines while exhibiting superior generalization across intersections in different cities. Moreover, the learned programmatic policies can be directly deployed on edge devices with minimal computational resources, further enhancing real-world applicability.

Cross-Domain Recommendation (CDR) can effectively alleviate the data sparsity issue in the recommendation system by transferring the source domain knowledge to the target domain. Many CDR methods try to find a mapping of latent embeddings of the overlapping nodes (e.g., the overlapping users) from source domains to target domains. However, due to the randomness of parameter initialization and model training, the embeddings from different domains may differ a lot and the inter-domain mapping can be difficult to learn. To this end, in this work, we propose a novel Similarity-Preserving Cross-Domain Recommendation ( SimCDR ) framework. Instead of trying to learn the inter-domain mapping, SimCDR aims to preserve the intra-domain similarities of users, such that the users that are close in the source domain are still close in the target domain. Specifically, we first verify that the intra-domain similarities of user embeddings can represent domain knowledge well. Then, based on this finding, we propose the SimCDR framework, which adds the auxiliary task of predicting similar users to preserve the intra-domain similarities of the source domain user embeddings, i.e., the source domain knowledge. Besides being distinct in knowledge transfer, SimCDR also supports multiple source domains and embedding compression, making it more suitable for real-world applications. We conduct thorough experiments that verify the effectiveness of the proposed method.

Sequential recommendation aims to predict the next item a user is likely to interact with based on their historical interaction sequence. Capturing user intent is crucial in this process, as each interaction is typically driven by specific intentions (e.g., buying skincare products for skin maintenance, buying makeup for cosmetic purposes, etc.). However, users often have multiple, dynamically changing intents, making it challenging for models to accurately learn these intents when relying on the entire historical sequence as input. To address this, we propose a novel framework called Intent Oriented Contrastive Learning for Sequential Recommendation (IOCLRec). This framework begins by segmenting users’ sequential behaviors into multiple subsequences, which represent the coarse-grained intents of users at different points in their interaction history. These subsequences form the basis for the three contrastive learning modules within IOCLRec. The fine-grained intent contrastive learning module uncovers detailed intent representations, while the single-intent and multi-intent contrastive learning modules utilize intent-oriented data augmentation operators to capture the diverse intents of users. These three modules work synergistically, driving comprehensive performance optimization in intricate sequential recommendation scenarios. Our method has been extensively evaluated on four public datasets, demonstrating superior effectiveness.

Jingyu Peng, Maolin Wang, Xiangyu Zhao, Kai Zhang, Wanyu Wang, Pengyue Jia, Qidong Liu, Ruocheng Guo, Qi Liu. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

Fine-grained sentiment analysis, which aims to identify sentiments associated with specific aspects within sentences, faces challenges in effectively incorporating commonsense knowledge. Recent advancements leveraging large language models (LLMs) as data generators show promise but are limited by the LLMs' lack of nuanced, domain-specific understanding and pose a significant risk of data leakage during inference, potentially leading to inflated performance metrics. To address these limitations, we propose LLM-Kit, a novel framework for commonsense-enhanced fine-grained sentiment analysis that integrates knowledge via LLM-guided graph construction, effectively mitigating data leakage risks. LLM-Kit operates in two key stages: (1) Commonsense Graph Construction (CGC): We design second-order rules and leverage LLMs for evaluation to ensure the accuracy of the generated graph and mitigate the risk of data leakage from LLMs. (2) Knowledge-integration Graph Representation Learning (KGRL): We extract knowledge that is aware of various aspects through Graph Representation Learning (GRL). To capture the underlying semantic nuances within the input sentence, we develop a Sentence Semantic Learning (SSL) module based on RoBERTa that explicitly encodes internal semantics. This module provides complementary information to the GCN, improving the model's ability to discern subtle sentiment variations related to different aspects. Comprehensive experiments on three public datasets affirm that LLM-Kit achieves comparable performance with state-of-the-art models.

Electrocardiogram (ECG) diagnosis aims to automatically classify ECG recordings into clinically meaningful categories, playing a vital role in medical decision-making. Deep learning methods, while promising, demand extensive annotated data and lack interpretability. Large Language Models (LLMs) offer potential in low-data scenarios and generating interpretable outputs, yet their application to ECG diagnosis, especially leveraging multimodal data (e.g., raw signals, derived features, and clinical knowledge), remains underexplored. To address these challenges, we propose ECG-Doctor, an interpretable and multimodal ECG diagnosis framework based on LLMs. ECG-Doctor comprises four key components: (1) ECG Knowledge Acquisition Module, which integrates external medical knowledge and Chain-of-Thought (CoT) reasoning to address the inability of LLMs to follow standardized ECG diagnostic procedures; (2) ECG Feature Extraction Module, which incorporates domain knowledge to overcome LLMs' limitations in comprehensively understanding structured ECG features; (3) ECG Waveform Analysis Module, which introduces time-series ECG models to equip LLMs with the capability to interpret and reason over raw ECG signal morphologies; (4) KNN-based ECG Retrieval Module, which retrieves the top-k most similar ECG samples and guides LLMs through in-context learning (ICL), enabling them to differentiate and learn from variations across ECGs. The outputs of these modules are aggregated and provided to the LLM as diagnostic context, enabling ICL to perform comprehensive ECG diagnosis. This design effectively simulates the diagnostic reasoning process of experienced electrocardiologists. Extensive experiments on the PTB-XL dataset demonstrate that ECG-Doctor is compatible with various LLMs and consistently outperforms existing baselines at both 100 Hz and 500 Hz sampling rates, showcasing its strong versatility and robustness. Furthermore, ECG-Doctor provides well-grounded diagnostic explanations, highlighting its superior interpretability.

As more and more internet users post images online to express their daily emotions, image sentiment analysis has attracted increasing attention. Recently, researchers generally tend to design different neural networks to extract visual features from images for sentiment analysis. Despite the significant progress, metadata, the data (e.g., text descriptions and keyword tags) for describing the image, has not been sufficiently explored in this task. In this paper, we propose a novel Metadata Enhanced Transformer for sentiment analysis (SentiFormer) to fuse multiple metadata and the corresponding image into a unified framework. Specifically, we first obtain multiple metadata of the image and unify the representations of diverse data. To adaptively learn the appropriate weights for each metadata, we then design an adaptive relevance learning module to highlight more effective information while suppressing weaker ones. Moreover, we further develop a cross-modal fusion module to fuse the adaptively learned representations and make the final prediction. Extensive experiments on three publicly available datasets demonstrate the superiority and rationality of our proposed method.

Graph In-Context Learning, with the ability to adapt pre-trained graph models to novel and diverse downstream graphs without updating any parameters, has gained much attention in the community. The key to graph in-context learning is to perform downstream graphs conditioned on chosen prompt examples. Existing methods randomly select subgraphs or edges as prompts, leading to noisy graph prompts and inferior model performance. Additionally, due to the gap between pre-training and testing graphs, when the number of classes in the testing graphs is much greater than that in the training, the in-context learning ability will also significantly deteriorate. To tackle the aforementioned challenges, we develop a multi-stage adaptive prompt optimization method GraphPrompter, which optimizes the entire process of generating, selecting, and using graph prompts for better in-context learning capabilities. Firstly, Prompt Generator introduces a reconstruction layer to highlight the most informative edges and reduce irrelevant noise for graph prompt construction. Furthermore, in the selection stage, Prompt Selector employs the k-nearest neighbors algorithm and pre-trained selection layers to dynamically choose appropriate sam-ples and minimize the influence of irrelevant prompts. Finally, we leverage a Prompt Augmenter with a cache replacement strategy to enhance the generalization capability of the pre-trained model on new datasets. Extensive experiments show that GraphPrompter effectively enhances the in-context learning ability of graph models. On average across all the settings, our approach surpasses the state-of-the-art baselines by over 8 %. Our code is released at https://ithub.com/karin0018/GraphPrompter.

Few-Shot Fake News Detection (FS-FND) aims to distinguish inaccurate news from real ones in extremely lowresource scenarios. This task has garnered increased attention due to the widespread dissemination and harmful impact of fake news on social media. Large Language Models (LLMs) have demonstrated competitive performance with the help of their rich prior knowledge and excellent in-context learning abilities. However, existing methods face significant limitations, such as the Understanding Ambiguity and Information Scarcity, which significantly undermine the potential of LLMs. To address these shortcomings, we propose a Dual-perspective Knowledge-guided Fake News Detection (DKFND) model, designed to enhance LLMs from both inside and outside perspectives. Specifically, DKFND first identifies the knowledge concepts of each news article through a Detection Module. Subsequently, DKFND creatively designs an Investigation Module to retrieve inside and outside valuable information concerning to the current news, followed by another Judge Module to evaluate the relevance and confidence of them. Finally, a Determination Module further derives two respective predictions and obtain the final result. Extensive experiments on two public datasets show the efficacy of our proposed method, particularly in low-resource settings.

Recent advances in machine learning, particularly in Natural Language Processing (NLP), have produced powerful models trained on vast datasets. However, these models risk leaking sensitive information, raising privacy concerns. In response, regulatory measures such as the European Union's General Data Protection Regulation (GDPR) have driven increasing interest in Machine Unlearning techniques, which enable models to selectively forget specific data entries. Early unlearning approaches primarily relied on pre-processing methods, while more recent research has shifted towards training-based solutions. Despite their effectiveness, a key limitation persists: most methods require access to original training data, which is often unavailable. Additionally, directly applying unlearning techniques bears the cost of undermining the model's expressive capabilities. To address these challenges, we introduce the Iterative Contrastive Unlearning (ICU) framework, which consists of three core components: A Knowledge Unlearning Induction module designed to target specific knowledge for removal using an unlearning loss; A Contrastive Learning Enhancement module to preserve the model's expressive capabilities against the pure unlearning goal; And an Iterative Unlearning Refinement module that dynamically adjusts the unlearning process through ongoing evaluation and updates. Experimental results demonstrate the efficacy of our ICU method in unlearning sensitive information while maintaining the model's overall performance, offering a promising solution for privacy-conscious machine learning applications.

Zero-shot composed image retrieval (ZS-CIR) is a rapidly growing area with significant practical applications, allowing users to retrieve a target image by providing a reference image and a relative caption describing the desired modifications. Existing ZS-CIR methods often struggle to capture fine-grained changes and integrate visual and semantic information effectively. They primarily rely on either transforming the multimodal query into a single text using image-to-text models or employing large language models for target image description generation, approaches that often fail to capture complementary visual information and complete semantic context. To address these limitations, we propose a novel Fine-Grained Zero-Shot Composed Image Retrieval method with Complementary VisualSemantic Integration (CVSI). Specifically, CVSI leverages three key components: (1) Visual Information Extraction, which not only extracts global image features but also uses a pre-trained mapping network to convert the image into a pseudo token, combining it with the modification text and the objects most likely to be added. (2) Semantic Information Extraction, which involves using a pre-trained captioning model to generate multiple captions for the reference image, followed by leveraging an LLM to generate the modified captions and the objects most likely to be added. (3) Complementary Information Retrieval, which integrates information extracted from both the query and database images to retrieve the target image, enabling the system to efficiently handle retrieval queries in a variety of situations. Extensive experiments on three public datasets (e.g., CIRR, CIRCO, and FashionIQ) demonstrate that CVSI significantly outperforms existing state-of-the-art methods. Our code is available at https://github.com/yyc6631/CVSI.

The advent of Large Language Models (LLMs) has revolutionized natural language processing, enabling advanced understanding and reasoning capabilities across a variety of tasks. Fine-tuning these models for specific domains, particularly through Parameter-Efficient Fine-Tuning (PEFT) strategies like LoRA, has become a prevalent practice due to its efficiency. However, this raises significant privacy and security concerns, as models may inadvertently retain and disseminate sensitive or undesirable information. To address these issues, we introduce a novel instance-wise unlearning framework, LLMEraser, which systematically categorizes unlearning tasks and applies precise parameter adjustments using influence functions. Unlike traditional unlearning techniques that are often limited in scope and require extensive retraining, LLMEraser is designed to handle a broad spectrum of unlearning tasks without compromising model performance. Extensive experiments on benchmark datasets demonstrate that LLMEraser excels in efficiently managing various unlearning scenarios while maintaining the overall integrity and efficacy of the models.

The reasoning abilities are one of the most enigmatic and captivating aspects of large language models (LLMs). Numerous studies are dedicated to exploring and expanding the boundaries of this reasoning capability. However, tasks that embody both reasoning and recall characteristics are often overlooked. In this paper, we introduce such a novel task, code reasoning, to provide a new perspective for the reasoning abilities of LLMs. We summarize three meta-benchmarks based on established forms of logical reasoning, and instantiate these into eight specific benchmark tasks. Our testing on these benchmarks reveals that LLMs continue to struggle with identifying satisfactory reasoning pathways. Additionally, we present a new pathway exploration pipeline inspired by human intricate problem-solving methods. This Reflective Hypothesis Decomposition and Amendment (RHDA) pipeline consists of the following iterative steps: (1) Proposing potential hypotheses based on observations and decomposing them; (2) Utilizing tools to validate hypotheses and reflection outcomes; (3) Revising hypothesis in light of observations. Our approach effectively mitigates logical chain collapses arising from forgetting or hallucination issues in multi-step reasoning, resulting in performance gains of up to $3\times$. Finally, we expanded this pipeline by applying it to simulate complex household tasks in real-world scenarios, specifically in VirtualHome, enhancing the handling of failure cases. We release our code and all of results at https://github.com/TnTWoW/code_reasoning.

Multimodal intent recognition (MIR) seeks to accurately interpret user intentions by integrating verbal and non-verbal information across video, audio and text modalities. While existing approaches prioritize text analysis, they often overlook the rich semantic content embedded in non-verbal cues. This paper presents a novel Wavelet-Driven Multimodal Intent Recognition (WDMIR) framework that enhances intent understanding through frequency-domain analysis of non-verbal information. To be more specific, we propose: (1) a wavelet-driven fusion module that performs synchronized decomposition and integration of video-audio features in the frequency domain, enabling fine-grained analysis of temporal dynamics; (2) a cross-modal interaction mechanism that facilitates progressive feature enhancement from bimodal to trimodal integration, effectively bridging the semantic gap between verbal and non-verbal information. Extensive experiments on MIntRec demonstrate that our approach achieves state-of-the-art performance, surpassing previous methods by 1.13% on accuracy. Ablation studies further verify that the wavelet-driven fusion module significantly improves the extraction of semantic information from non-verbal sources, with a 0.41% increase in recognition accuracy when analyzing subtle emotional cues.

With the rapid progress of large language models (LLMs) and diffusion models, there has been growing interest in personalized content generation. However, current conversational systems often present the same recommended content to all users, falling into the dilemma of "one-size-fits-all." To break this limitation and boost user engagement, in this paper, we introduce PCG (**P**ersonalized Visual **C**ontent **G**eneration), a unified framework for personalizing item images within conversational systems. We tackle two key bottlenecks: the depth of personalization and the fidelity of generated images. Specifically, an LLM-powered Inclinations Analyzer is adopted to capture user likes and dislikes from context to construct personalized prompts. Moreover, we design a dual-stage LoRA mechanism—Global LoRA for understanding task-specific visual style, and Local LoRA for capturing preferred visual elements from conversation history. During training, we introduce the visual content condition method to ensure LoRA learns both historical visual context and maintains fidelity to the original item images. Extensive experiments on benchmark conversational datasets—including objective metrics and GPT-based evaluations—demonstrate that our framework outperforms strong baselines, which highlight its potential to redefine personalization in visual content generation for conversational scenarios like e-commerce and real-world recommendation.

Modern recommender systems face a critical challenge in modeling the intricate interplay between multi-behavior interactions of users (e.g., clicks, adds-to-cart and purchases) and temporal dynamics that drive evolving preferences. While existing multi-behavior sequential recommendation methods attempt to capture these signals, they often suffer from fragmented modeling, such as decoupling behaviors and items into separate sequences, neglecting time-aware transitions, or relying on computationally intensive architectures that hinder real-world scalability. To address these limitations, we propose GEneralized Alternating Regressor (GEAR), a novel framework that unifies behaviors, items, and temporal contexts into a single autoregressive sequence through an alternating architecture. At its core, GEAR represents user interactions as triplets and processes them through a modular transformer architecture. In this architecture, each triplet is alternately modeled at lower layers to disentangle fine-grained patterns, while upper layers jointly learn cross-signal dependencies. This design mimics the interlocking mechanism of gears, enabling the seamless transitions between multi-behavior dynamics and item transitions. Additionally, we incorporate a time-bias term to quantify the decay of behavioral influence across both short- and long-term horizons. Extensive experiments on real-world datasets validate the effectiveness, generalizability, and computational efficiency of the proposed framework.

Transmission Control Protocol (TCP) congestion control is a fundamental mechanism in the Internet that maintains network stability and performance by adjusting the sending rate of connections. Recently, Deep Reinforcement Learning (DRL) methods have shown superior performance over traditional expert-designed solutions. However, the DRL policies are often represented by black-box neural networks, they lack interpretability, making verification challenging and requiring excessive floating-point computation. This work introduces a novel approach, Programmatic reinforcement learning for Congestion Control (ProCC), designed to autonomously discover a program as a control policy from scratch. Programs in ProCC include branching structures (e.g., if blocks and if-else blocks), conditions and actions. However, directly optimizing such program structures is challenging due to their discrete non-differentiable nature, and the program space grows exponentially as the depth increases. To address this issue, ProCC defines a Domain-Specific Language (DSL) and program transformation rules, enabling the construction of a program search graph where similar programs are closer in proximity. Subsequently, ProCC employs Monte Carlo Tree Search (MCTS) to efficiently explore the discrete space and obtain promising programs. Extensive experiments conducted in multiple simulated environments demonstrate that ProCC is adaptive and consistently performs well under varying network conditions. The learned program's performance surpasses that of state-of-the-art DRL agents, and more importantly, the generated policies are concise, transparent, and computationally efficient.

Online learning has demonstrated superiority in connecting high-quality educational resources to a global audience. To ensure an excellent learning experience with sustainable and opportune learning instructions, online learning systems must comprehend learners' evolving knowledge states based on their learning interactions, known as the Knowledge Tracing (KT) task. Generally, learners practice through various quizzes, each comprising several exercises that cover similar knowledge concepts. Therefore, their learning interactions are continuous within each quiz but discrete across different quizzes. However, existing methods overlook the quiz structure and assume all learning interactions are uniformly distributed. We argue that learners' knowledge states should also be assessed in quiz since they practiced in quiz. To achieve this goal, we present a novel Quiz-based Knowledge Tracing (QKT) model, which effectively integrates the quiz structure of learning interactions. This is achieved by designing two distinct modules by neural networks: one for intra-quiz modeling and another for inter-quiz fusion. Extensive experimental results on public real-world datasets demonstrate that QKT achieves new state-of-the-art performance. The findings of this study suggest that incorporating the quiz structure of learning interactions can efficiently comprehend learners' knowledge states with fewer quizzes, and provides valuable insights into designing effective quizzes with fewer exercises.

Due to cultural influences and the long history of language evolution, metaphorical expressions are pervasive in Chinese texts, particularly in ironical comments, poetry, and various literary works. Traditional Chinese text classification methods relying on surface-level features often fail to bridge the gap between literal features and label spaces in such texts. To address this problem, we propose L-MAM, a novel label-constrained contrastive learning approach for Chinese text classification based on large language models. Specifically, we first introduce an ambiguity recognition module (ARM) to quantify phonetic and syntactic ambiguities in Chinese characters, identifying expressions that benefit most from metaphorical associations. By leveraging large language model-driven prompts, metaphorical interpretations are extracted from texts, and classification labels are aligned with semantic definitions for deeper contextual understanding. Then, we design a contextual attention mechanism that dynamically adjusts weights based on character ambiguity, and a metaphorical attention mechanism that aligns metaphorical embeddings with label semantics for refined label-constrained associations. Additionally, we devise a contextual-metaphorical contrastive learning mechanism, which employs a triplet loss to differentiate literal and metaphorical aspects while enhancing interclass separability. Finally, we conduct extensive experiments on three real-world datasets, where the experimental results validate the effectiveness of L-MAM, offering new insights into Chinese text classification involving metaphorical comprehension.

Cognitive diagnosis (CD) models assess students’ knowledge state by analyzing test performance and identifying specific knowledge concepts mastered. However, traditional CD models often overlook dynamic changes in knowledge state during exercise and exam. To address this, we proposetracking knowledge state transitions in cognitive diagnosis (TKT-CD). The TKT-CD model introduces two key modules: the time-driven knowledge state transition (TDKT) module and the cognition-driven knowledge state transition (CDKT) module. TDKT captures temporal transitions in knowledge state during exercise, while CDKT analyzes relationships between knowledge concepts during exercise and exam to enhance reasoning accuracy. Combining neural networks with traditional static CD methods, TKT-CD resolves inconsistencies in knowledge state and provides more accurate representations of student interactions. Empirical studies show that TKT-CD significantly outperforms existing CD methods in predicting student performance, particularly in exercise and exam-based learning modes, offering a more effective tool for personalized education.

Sequential recommendation systems are integral to discerning temporal user preferences. Yet, the task of learning from abbreviated user interaction sequences poses a notable challenge. Data augmentation has been identified as a potent strategy to enhance the informational richness of these sequences. Traditional augmentation techniques, such as item randomization, may disrupt the inherent temporal dynamics. Although recent advancements in reverse chronological pseudo-item generation have shown promise, they can introduce temporal discrepancies when assessed in a natural chronological context. In response, we introduce a sophisticated approach, Bidirectional temporal data Augmentation with pre-training (BARec). Our approach leverages bidirectional temporal augmentation and knowledge-enhanced fine-tuning to synthesize authentic pseudo-prior items thatretain user preferences and capture deeper item semantic correlations, thus boosting the model’s expressive power. Our comprehensive experimental analysis on five benchmark datasets confirms the superiority of BARec across both short and elongated sequence contexts. Moreover, theoretical examination and case study offer further insight into the model’s logical processes and interpretability.

Knowledge Tracing (KT) predicts future performance by modeling students' historical interactions, and understanding students' affective states can enhance the effectiveness of KT, thereby improving the quality of education. Although traditional KT values students' cognition and learning behaviors, efficient evaluation of students' affective states and their application in KT still require further exploration due to the non-affect-oriented nature of the data and budget constraints. To address this issue, we propose a computation-driven approach,DynamicAffectSimulationKnowledgeTracing (DASKT), to explore the impact of various student affective states (such as frustration, concentration, boredom, and confusion) on their knowledge states. In this model, we first extract affective factors from students' non-affect-oriented behavioral data, then use clustering and spatiotemporal sequence modeling to accurately simulate students' dynamic affect changes when dealing with different problems. Subsequently, we incorporate affect with time-series analysis to improve the model's ability to infer knowledge states over time and space. Extensive experimental results on two public real-world educational datasets show that DASKT can achieve more reasonable knowledge states under the effect of students' affective states. Moreover, DASKT outperforms the most advanced KT methods in predicting student performance. Our research highlights a promising avenue for future KT studies, focusing on achieving high interpretability and accuracy.

Code summarization is designed to generate descriptive natural language for code snippets, facilitating understanding and increasing productivity for developers. Previous research often overlooks the semantic connection between code and its natural language description, resulting in a noticeable gap and suboptimal solution. To address this issue, we introduce a semantic-aligned code summarization framework that leverages crucial data flow information from code for semantic analysis, ensuring alignment between code and summaries. Specifically, we utilize a semantic extraction module (SEM) to decipher the meaning of code and align it with natural language through a semantic alignment module. In the SEM, we construct a code graph that includes data flow edges using static program analysis techniques. Then, on this well-constructed code graph, we innovatively adopt a walking algorithm guided by data flow to extract the semantics of the code. This walking algorithm understands code semantics by analyzing the information transfer between variables during the program execution process. In the semantic alignment module, we integrate a contrastive learning loss mechanism for semantic alignment, which cohesively maps the semantic domains of code and natural language into a unified vector space. We further theoretically analyzed that the data-flow-guided walking algorithm can ensure capturing semantically highly related nodes in shorter paths. Extensive experiments on two benchmark datasets demonstrate the efficacy and broad applicability of the framework.

The recent advancements in Deep Reinforcement Learning (DRL) have significantly enhanced the performance of adaptive Traffic Signal Control (TSC). However, DRL policies are typically represented by neural networks, which are over-parameterized black-box models. As a result, the learned policies often lack interpretability, and cannot be deployed directly in the real-world edge hardware due to resource constraints. In addition, the DRL methods often exhibit limited generalization performance, struggling to generalize the learned policy to other geographical regions. These factors limit the practical application of learning-based approaches. To address these issues, we suggest the use of an inherently interpretable program for representing the control policy. We present a new approach, Programmatic Interpretable reinforcement learning for traffic signal control (π-light), designed to autonomously discover non-differentiable programs. Specifically, we define a Domain Specific Language (DSL) and transformation rules for constructing programs, and utilize Monte Carlo Tree Search (MCTS) to find the optimal program in a discrete space. Extensive experiments demonstrate that our method consistently outperforms baseline approaches. Moreover, π-Light exhibits superior generalization capabilities compared to DRL, enabling training and evaluation across intersections from different cities. Finally, we analyze how the learned program policies can directly deploy on edge devices with extremely limited resources.

Zero-Shot Relational Learning (ZSRL), strives to predict relations that have not been observed during training, presenting a considerable challenge in terms of model generalization. Existing ZSRL methods usually utilize the prior knowledge of labels (e.g., text description, ontological schema) to enable knowledge transfer by learned features. Nonetheless, these methods remain limited to calculating the surface features exhibited by relations, failing to fully explore their underlying driving factors. This leads to insufficient discrimination between the shared and distinctive inherent components among relations, which consequently impedes the cognitive understanding required for advanced reasoning. In our study, we aim to identify and utilize shared factors that widely exist in the prior knowledge of classes to learn enhanced semantic representations via shared factors composition, and develop our Factor-based ZSRL framework (FZR) with Generative Adversarial Networks (GANs) to bridge inequality between seen and unseen classes. FZR is designed to restructure the semantic space in such a way that it captures the essence of relation formation, thereby facilitating superior knowledge transfer in zero-shot scenarios. We conduct extensive experiments and evaluate our model on real-world datasets, and the results clearly demonstrate the effectiveness of the proposed model in zero-shot relational learning tasks.

Aspect-based sentiment analysis (ABSA) aims to predict the sentiment polarity of a specific aspect within a given sentence.Most existing methods predominantly leverage semantic or syntactic information based on attention scores, which are susceptible to interference caused by irrelevant contexts and often lack sentiment knowledge at a data-specific level.In this paper, we propose a novel Dynamic Multigranularity Attribution Network (DMAN) from the perspective of attribution.Initially, we leverage Integrated Gradients to dynamically extract attribution scores for each token, which contain underlying reasoning knowledge for sentiment analysis.Subsequently, we aggregate attribution representations from multiple semantic granularities in natural language, enhancing a profound understanding of the semantics.Finally, we integrate attribution scores with syntactic information to capture the relationships between aspects and their relevant contexts more accurately during the sentence understanding process.Extensive experiments on five benchmark datasets demonstrate the effectiveness of our proposed method.

Entity Linking (EL) is the process of associating ambiguous textual mentions to specific entities in a knowledge base.Traditional EL methods heavily rely on large datasets to enhance their performance, a dependency that becomes problematic in the context of few-shot entity linking, where only a limited number of examples are available for training.To address this challenge, we present OneNet, an innovative framework that utilizes the few-shot learning capabilities of Large Language Models (LLMs) without the need for fine-tuning.To the best of our knowledge, this marks a pioneering approach to applying LLMs to few-shot entity linking tasks.OneNet is structured around three key components prompted by LLMs: (1) an entity reduction processor that simplifies inputs by summarizing and filtering out irrelevant entities, (2) a dual-perspective entity linker that combines contextual cues and prior knowledge for precise entity linking, and (3) an entity consensus judger that employs a unique consistency algorithm to alleviate the hallucination in the entity linking reasoning.Comprehensive evaluations across seven benchmark datasets reveal that OneNet outperforms current stateof-the-art entity linking methods.

Chinese Spelling Check (CSC) aims to identify and correct spelling errors in Chinese texts, where enhanced semantic understanding of a sentence can significantly improve correction accuracy.Recently, Large Language Models (LLMs) have demonstrated exceptional mastery of world knowledge and semantic understanding, rendering them more robust against spelling errors.However, the application of LLMs in CSC is a double-edged sword, as they tend to unnecessarily alter sentence length and modify rare but correctly used phrases.In this paper, by leveraging the capabilities of LLMs while mitigating their limitations, we propose a novel plug-and-play Alignment-and-Replacement Module (ARM) that enhances the performance of existing CSC models and without the need for retraining or fine-tuning.Experiment results and analysis on three benchmark datasets demonstrate the effectiveness and competitiveness of the proposed module.

Large Language Models (LLMs) have shown remarkable capabilities in various natural language processing tasks.However, LLMs may rely on dataset biases as shortcuts for prediction, which can significantly impair their robustness and generalization capabilities.This paper presents Shortcut Suite, a comprehensive test suite designed to evaluate the impact of shortcuts on LLMs' performance, incorporating six shortcut types, five evaluation metrics, and four prompting strategies.Our extensive experiments yield several key findings: 1) LLMs demonstrate varying reliance on shortcuts for downstream tasks, significantly impairing their performance.2) Larger LLMs are more likely to utilize shortcuts under zero-shot and few-shot in-context learning prompts.3) Chain-of-thought prompting notably reduces shortcut reliance and outperforms other prompting strategies, while fewshot prompts generally underperform compared to zero-shot prompts.4) LLMs often exhibit overconfidence in their predictions, especially when dealing with datasets that contain shortcuts.5) LLMs generally have a lower explanation quality in shortcut-laden datasets, with errors falling into three types: distraction, disguised comprehension, and logical fallacy.Our findings offer new insights for evaluating robustness and generalization in LLMs and suggest potential directions for mitigating the reliance on shortcuts.The code is available at https://github.com/ yyhappier/ShortcutSuite.git.

Graph Neural Networks (GNNs) have demonstrated remarkable performance in graph classification tasks. However, ensuring the explainability of their predictions remains a challenge. To address this, graph rationalization methods have been introduced to generate concise subsets of the original graph, known as rationales, which serve to explain the predictions made by GNNs. Existing rationalizations often rely on shortcuts in data for prediction and rationale composition. In response, de-shortcut rationalization methods have been proposed, which commonly leverage counterfactual augmentation to enhance data diversity for mitigating the shortcut problem. Nevertheless, these methods have predominantly focused on centralized datasets and have not been extensively explored in the Federated Learning (FL) scenarios. To this end, in this paper, we propose a Federated Graph Rationalization (FedGR) with anti-shortcut augmentations to achieve self-explaining GNNs, which involves two data augmenters. These augmenters are employed to produce client-specific shortcut conflicted samples at each client, which contributes to mitigating the shortcut problem under the FL scenarios. Experiments on real-world benchmarks and synthetic datasets validate the effectiveness of FedGR under the FL scenarios.

Molecular docking, a technique for predicting ligand binding poses, is crucial in structure-based drug design for understanding protein-ligand interactions. Recent advancements in docking methods, particularly those leveraging geometric deep learning (GDL), have demonstrated significant efficiency and accuracy advantages over traditional sampling methods. Despite these advancements, current methods are often tailored for specific docking settings, and limitations such as the neglect of protein side-chain structures, difficulties in handling large binding pockets, and challenges in predicting physically valid structures exist. To accommodate various docking settings and achieve accurate, efficient, and physically reliable docking, we propose a novel two-stage docking framework, DeltaDock, consisting of pocket prediction and site-specific docking. We innovatively reframe the pocket prediction task as a pocket-ligand alignment problem rather than direct prediction in the first stage. Then we follow a bi-level coarse-to-fine iterative refinement process to perform site-specific docking. Comprehensive experiments demonstrate the superior performance of DeltaDock. Notably, in the blind docking setting, DeltaDock achieves a 31\% relative improvement over the docking success rate compared with the previous state-of-the-art GDL model DiffDock. With the consideration of physical validity, this improvement increases to about 300\%.

Formula recognition endeavors to automatically identify mathematical formulas from images. Currently, the Encoder-Decoder model has significantly advanced the translation from image to corresponding formula markups. Nonetheless, previous research primarily concentrated on single-line formula recognition, ignoring the recognition of multi-line formulas, which presents additional challenges such as more stringent grammatical restrictions and two- dimensional positions. In this work, we present GAP (Grammar And Position-Aware formula recognition), a comprehensive framework designed to tackle the challenges in multi-line mathematical formula recognition. First, to overcome the limitations imposed by grammar, we design a novel Grammar Aware Contrastive Learning (GACL) module, integrating complex grammar rules into the transcription model through a contrastive learning mechanism. Furthermore, primitive contrastive learning lacks clear directions for comprehending grammar rules and can lead to unstable convergence or prolonged training cycles. To enhance training efficiency, we propose Rank-Based Sampling (RBS) specialized for multi-line formulas, which guides the learning process by the importance ranking of different grammar errors. Finally, spatial location information is critical considering the two-dimensional nature of multi-line formulas. To aid the model in keeping track of that global information, we introduced a Visual Coverage (VC) mechanism that incorporates historical attention information into the image features via a parameter-free way. To validate the effectiveness of our GAP framework, we construct a new dataset Multi-Line containing 12,002 multi-line formulas and conduct extensive experiments to show the efficacy of our GAP framework in capturing grammatical rules, enhancing recognition accuracy, and enhancing training efficiency. Codes and datasets are available at https://github.com/Sinon02/GAP.

Recommendation systems have attracted attention from academia and industry due to their wide range of application scenarios. However, cold start remains a challenging problem limited by sparse user interactions. Some scholars propose to transfer the dense information from the source domain to the target domain through cross-domain recommendation, but most of the work assumes that there is a small amount of historical interaction in the target domain. However, this approach essentially presupposes the existence of at least some historical interaction within the target domain. In this paper, we focus on the domain-level zero-shot recommendation (DZSR) problem. To address the above challenges, we propose a knowledge-aware cross-semantic alignment (K-CSA) framework to learn transferable source domain semantic information. The motivation is to establish stable alignments of interests in different domains through class semantic descriptions (CSDs). Specifically, due to the lack of effective information in the target domain, we learn semantic representations of source and target domain items based on knowledge graphs. Moreover, we conduct multi-view K-means to extract item CSDs from the learned semantic representations. Further, K-CSA learns universal user CSDs through the designed multi-head self-attention. To facilitate the transference of user interest from the source domain to the target domain, we devise a cross-semantic contrastive learning strategy, grounded in the prototype distribution matrix. We conduct extensive experiments on several real-world cross-domain datasets, and the experimental results clearly demonstrate the superiority of our proposed K-CSA compared with other baselines.

Commonsense Question Answering (CQA) aims to answer questions that require human commonsense. Closed-book CQA, as one of the subtasks, requires the model to answer questions without retrieving external knowledge, which emphasizes the importance of the model's problem-solving ability. Most previous methods relied on large-scale pre-trained models to generate question-related knowledge while ignoring the crucial role of skills in the process of answering commonsense questions. Generally, skills refer to the learned ability in performing a specific task or activity, which are derived from knowledge and experience. In this paper, we introduce a new approach named Dynamic Skill-aware Commonsense Question Answering (DSCQA), which transcends the limitations of traditional methods by informing the model about the need for each skill in questions and utilizes skills as a critical driver in CQA process. To be specific, DSCQA first employs commonsense skill extraction module to generate various skill representations. Then, DSCQA utilizes dynamic skill module to generate dynamic skill representations. Finally, in perception and emphasis module, various skills and dynamic skill representations are used to help question-answering process. Experimental results on two publicly available CQA datasets show the effectiveness of our proposed model and the considerable impact of introducing skills.

User modeling, which aims to capture users' characteristics or interests, heavily relies on task-specific labeled data and suffers from the data sparsity issue. Several recent studies tackled this problem by pre-training the user model on massive user behavior sequences with a contrastive learning task. Generally, these methods assume different views of the same behavior sequence constructed via data augmentation are semantically consistent, i.e., reflecting similar characteristics or interests of the user, and thus maximizing their agreement in the feature space. However, due to the diverse interests and heavy noise in user behaviors, existing augmentation methods tend to lose certain characteristics of the user or introduce noisy behaviors. Thus, forcing the user model to directly maximize the similarity between the augmented views may result in a negative transfer. To this end, we propose to replace the contrastive learning task with a new pretext task: Augmentation-Adaptive SelfSupervised Ranking (AdaptSSR), which alleviates the requirement of semantic consistency between the augmented views while pre-training a discriminative user model. Specifically, we adopt a multiple pairwise ranking loss which trains the user model to capture the similarity orders between the implicitly augmented view, the explicitly augmented view, and views from other users. We further employ an in-batch hard negative sampling strategy to facilitate model training. Moreover, considering the distinct impacts of data augmentation on different behavior sequences, we design an augmentation-adaptive fusion mechanism to automatically adjust the similarity order constraint applied to each sample based on the estimated similarity between the augmented views. Extensive experiments on both public and industrial datasets with six downstream tasks verify the effectiveness of AdaptSSR.

Automatically generating controllable and diverse mathematical word problems (MWPs) which conform to equations and topics is a crucial task in information retrieval and natural language generation. Recent deep learning models mainly focus on improving the problem readability but overlook the mathematical logic coherence, which tends to generate unsolvable problems. In this paper, we draw inspiration from the human problem-designing process and propose a Mathematical structure Planning and Knowledge enhanced Generation model (MaPKG), following the "plan-then-generate" steps. Specifically, we propose a novel dynamic planning module to make sentence-level equation plans and a dual-attention mechanism for word-level generation, incorporating equation structure representation and external commonsense knowledge. Extensive experiments on two MWP datasets show our model can guarantee more solvable, high-quality, and diverse problems. Our code is available at https://github.com/KenelmQLH/MaPKG.git

In online games, predicting massive battle outcomes is a fundamental task of many applications, such as team optimization and tactical formulation. Existing works do not pay adequate attention to the massive battle. They either seek to evaluate individuals in isolation or mine simple pair-wise interactions between individuals, neither of which effectively captures the intricate interactions between massive units (e.g., individuals). Furthermore, as the team size increases, the phenomenon of diminishing marginal utility of units emerges. Such a diminishing pattern is rarely noticed in previous work, and how to capture it from data remains a challenge. To this end, we propose a novel Massive battle outcome predictor with margiNal Effect modules, namely MassNE, which comprehensively incorporates individual effects, cooperation effects (i.e., intra-team interactions) and suppression effects (i.e., inter-team interactions) for predicting battle outcomes. Specifically, we design marginal effect modules to learn how units’ marginal utility changing respect to their number, where the monotonicity assumption is applied to ensure rationality. In addition, we evaluate the current classical models and provide mathematical proofs that MassNE is able to generalize several earlier works in massive settings. Massive battle datasets generated by StarCraft II APIs are adopted to evaluate the performances of MassNE. Extensive experiments empirically demonstrate the effectiveness of MassNE, and MassNE can reveal reasonable cooperation effects, suppression effects, and marginal utilities of combat units from the data.

Aspect-level sentiment analysis is a granular emotional classification task that refers to identifying sentiment polarities towards aspects in a sentence. Although previous research has reached a great achievement, this task remains very challenging. First, previous approaches only focus on one specific domain, which lacks the capability of transferring to other domains. Moreover, the majority of prior studies ignore the direct relationship between aspects and the corresponding sentiment words. To this end, in this paper, we propose a novel model named Efficient Adaptive Transfer Network (EATN) for aspect-level sentiment analysis which emphasizes the need of incorporating the correlation among multiple domains. The proposed EATN provides a Domain Adaptation Module (DAM) to learn common features from the sufficiently labeled source domain and to guide the classification performance in the target domain. Specifically, DAM comprises two special tasks, with one sentiment classification task aiming to learn sentiment knowledge and the other domain classification task focusing on learning domain-invariant features. Moreover, we design aspect-aware multi-head attention mechanism to capture the direct associations between the aspects and the contextual sentiment words, which is beneficial to learn the aspect-aware semantic knowledge. Extensive experiments demonstrate the effectiveness and universality of our method.

Pre-trained Language Models (PLMs) have achieved remarkable performance gains across numerous downstream tasks in natural language understanding. Various Chinese PLMs have been successively proposed for learning better Chinese language representation. However, most current models use Chinese characters as inputs and are not able to encode semantic information contained in Chinese words. While recent pre-trained models incorporate both words and characters simultaneously, they usually suffer from deficient semantic interactions and fail to capture the semantic relation between words and characters. To address the above issues, we propose a simple yet effective PLM CLOWER, which adopts the Contrastive Learning Over Word and charactER representations. In particular, CLOWER implicitly encodes the coarse-grained information (i.e., words) into the fine-grained representations (i.e., characters) through contrastive learning on multi-grained information. CLOWER is of great value in realistic scenarios since it can be easily incorporated into any existing fine-grained based PLMs without modifying the production pipelines. Extensive experiments conducted on a range of downstream tasks demonstrate the superior performance of CLOWER over several state-of-the-art baselines.

Cross-domain sentiment classification (CDSC) aims to use the transferable semantics learned from the source domain to predict the sentiment of reviews in the unlabeled target domain. Existing studies in this task attach more attention to the sequence modeling of sentences while largely ignoring the rich domain-invariant semantics embedded in graph structures (i.e., the part-of-speech tags and dependency relations). As an important aspect of exploring characteristics of language comprehension, adaptive graph representations have played an essential role in recent years. To this end, in the paper, we aim to explore the possibility of learning invariant semantic features from graph-like structures in CDSC. Specifically, we present Graph Adaptive Semantic Transfer (GAST) model, an adaptive syntactic graph embedding method that is able to learn domain-invariant semantics from both word sequences and syntactic graphs. More specifically, we first raise a POS-Transformer module to extract sequential semantic features from the word sequences as well as the part-of-speech tags. Then, we design a Hybrid Graph Attention (HGAT) module to generate syntax-based semantic features by considering the transferable dependency relations. Finally, we devise an Integrated aDaptive Strategy (IDS) to guide the joint learning process of both modules. Extensive experiments on four public datasets indicate that GAST achieves comparable effectiveness to a range of state-of-the-art models.

Knowledge tracing is a crucial and fundamental task in online education systems, which can predict students' knowledge state for personalized learning. Unfortunately, existing methods are domain-specific, whereas there are many domains (e.g., subjects, schools) in the real education scene and some domains suffer from the problem of lacking sufficient data. Therefore, how to exploit the knowledge in other domains, to improve the model's performance for target domain remains pretty much open. We term this problem as Domain Adaptation for Knowledge Tracing (DAKT), which aims to transfer knowledge from the source domain to the target one for knowledge tracing. In this paper, we propose a novel adaptable method, namely Adaptable Knowledge Tracing (AdaptKT), which contains three phases to explore this problem. Specifically, phase I is instance selection. Given the question texts of two domains, we train an auto-encoder to select and embed similar instances from both domains. Phase II is distribution discrepancy minimizing. After obtaining the selected instances and their linguistic representations, we train a knowledge tracing model and adopt the Maximum Mean Discrepancy (MMD) to minimize the discrepancy between the distributions of the domain-specific knowledge states. Phase III is fine-tuning of the output layer. We replace the output layer of the model that trained in phase II by a new one to make the knowledge tracing model's output dimension matches the number of knowledge concepts in the target domain. The new output layer is trained while other parameters that before it are frozen. We conduct extensive experiments on two large-scale real-world datasets, where the experimental results clearly demonstrate the effectiveness of AdaptKT for solving DAKT problem. We will public the code on the Github after the acceptance of the paper.

Modern recommendation systems introduce the re-ranking stage to optimize the entire list directly. This paper focuses on the design of re-ranking framework in feed to optimally model the mutual influence between items and further promote user engagement. On mobile devices, users browse the feed almost in a top-down manner and rarely compare items back and forth. Besides, users often compare item with its adjacency based on their partial observations. Given the distinct user behavior patterns, the modeling of mutual influence between items should be carefully designed. Existing re-ranking models encode the mutual influence between items with sequential encoding methods. However, previous works may be dissatisfactory due to the ignorance of connections between items on different scopes. In this paper, we first discuss Unidirectivity and Locality on the impacts and consequences, then report corresponding solutions in industrial applications. We propose a novel framework based on the empirical evidence from user analysis. To address the above problems, we design a \underlineS cope-aware \underlineR e-ranking with \underlineG ated \underlineA ttention model (SRGA ) to emulate the user behavior patterns from two aspects: 1) we emphasize the influence along the user's common browsing direction; 2) we strength the impacts of pivotal adjacent items within the user visual window. Specifically, we design a global scope attention to encode inter-item patterns unidirectionally from top to bottom. Besides, we devise a local scope attention sliding over the recommendation list to underline interactions among neighboring items. Furthermore, we design a learned gate mechanism to aggregating the information dynamically from local and global scope attention. Extensive offline experiments and online A/B testing demonstrate the benefits of our novel framework. The proposed SRGA model achieves the best performance in offline metrics compared with the state-of-the-art re-ranking methods. Further, empirical results on live traffic validate that our recommender system, equipped with SRGA in the re-ranking stage, improves significantly in user engagement.

Match outcome prediction in group comparison setting is a challenging but important task. Existing works mainly focus on learning individual effects or mining limited interactions between teammates, which is not sufficient for capturing complex interactions between teammates as well as between opponents. Besides, the importance of interacting with different characters is still largely underexplored. To this end, we propose a novel Neural Attentional Cooperation-competition model (NeuralAC), which incorporates weighted-cooperation effects (i.e., intra-team interactions) and weighted-competition effects (i.e., inter-team interactions) for predicting match outcomes. Specifically, we first project individuals to latent vectors and learn complex interactions through deep neural networks. Then, we design two novel attention-based mechanisms to capture the importance of intra-team and inter-team interactions, which enhance NeuralAC with both accuracy and interpretability. Furthermore, we demonstrate NeuralAC can generalize several previous works. To evaluate the performances of NeuralAC, we conduct extensive experiments on four E-sports datasets. The experimental results clearly verify the effectiveness of NeuralAC compared with several state-of-the-art methods.

Recent studies in recommender systems have managed to achieve significantly improved performance. However, despite being extensively studied, these methods still suffer from two limitations. First, previous studies either encode the document or extract latent sentiment via neural networks, which are difficult to interpret the sentiment of reviewers intuitively. Second, they neglect the personalized interaction of reviews with user/item, i.e., each review has different contributions when modeling the preference of user/item

Assessing the proficiency of trial lawyers in different legal fields is of significant importance since a qualified lawyer or lawyer team can strive for his clients' best rights while ensuring the fairness of litigations. However, proficiency assessment for lawyers is very challenging due to many technical and domain challenges, such as the lack of unified evaluation standards, and the complex interactions between lawyers and cases in real legal systems. To this end, we propose a novel proficiency assessment network for trial lawyers (LawyerPAN) to quantify lawyer proficiency through online litigation records. Specifically, we first leverage the theories in psychological measurement for mapping the proficiency of lawyers in each field into a unified real number space. Meanwhile, the characteristics of cases (i.e., case difficulty and discrimination) are well modeled to ensure fairness when assessing lawyers in different cases and fields. Then, we model the interactions between lawyers and cases from two perspectives: the anticipatory perspective aims to measure the personal proficiency of anticipated strategy, and the adversarial perspective seeks to depict the gap of lawyers' proficiency between both sides (i.e., plaintiffs and defendants). Finally, we conduct extensive experiments on real-world data, and the results show the effectiveness and interpretability of our approaches on assessing the proficiency of trial lawyers.

Legal Judgment Prediction is a fundamental task in legal intelligence of the civil law system, which aims to automatically predict the judgment results of multiple subtasks, such as charge, law article, and term of penalty prediction. Existing studies mainly focus on the impact of the entire fact description on all subtasks. They ignore the practical judicial scenario, where judges adopt circumstances of crime (i.e., various parts of the fact) to decide judgment results. To this end, in this paper, we propose a circumstance-aware legal judgment prediction framework (i.e., NeurJudge) by exploring circumstances of crime. Specifically, NeurJudge utilizes the results of intermediate subtasks to separate the fact description into different circumstances and exploits them to make the predictions of other subtasks. In addition, considering the popularity of confusing verdicts (i.e., charges and law articles), we further extend NeurJudge to a more comprehensive framework which is denoted by NeurJudge+. Particularly, NeurJudge+ utilizes a label embedding method to incorporate the semantics of labels (i.e., charges and law articles) into facts to generate more expressive fact representations for confusing verdicts problems. Extensive experimental results on two real-world datasets clearly validate the effectiveness of our proposed frameworks.

In the Click-Through Rate (CTR) prediction scenario, user's sequential behaviors are well utilized to capture the user interest in the recent literature. However, despite being extensively studied, these sequential methods still suffer from three limitations. First, existing methods mostly utilize attention on the behavior of users, which is not always suitable for CTR prediction, because users often click on new products that are irrelevant to any historical behaviors. Second, in the real scenario, there are numerous users that have operations a long time ago, but turn relatively inactive in recent times. Thus, it is hard to precisely capture user's current preferences through early behaviors. Third, multiple representations of user's historical behaviors in different feature subspaces are largely ignored. To remedy these issues, we propose a Multi-Interactive Attention Network (MIAN) to comprehensively extract the latent relationship among all kinds of fine-grained features (e.g., gender, age and occupation in user-profile). Specifically, MIAN contains a Multi-Interactive Layer (MIL) that integrates three local interaction modules to capture multiple representations of user preference through sequential behaviors and simultaneously utilize the fine-grained user-specific as well as context information. In addition, we design a Global Interaction Module (GIM) to learn the high-order interactions and balance the different impacts of multiple features. Finally, Offline experiment results from three datasets, together with an Online A/B test in a large-scale recommendation system, demonstrate the effectiveness of our proposed approach.

Cross-domain sentiment classification refers to utilizing useful knowledge in the source domain to help sentiment classification in the target domain which has few or no labeled data. Most existing methods mainly concentrate on extracting common features between domains. Unfortunately, they cannot fully consider the effects of the aspect (e.g., the battery life in reviewing an electronic product) information of the sentences. In order to better solve this problem, we propose an Interactive Attention Transfer Network (IATN) for crossdomain sentiment classification. IATN provides an interactive attention transfer mechanism, which can better transfer sentiment across domains by incorporating information of both sentences and aspects. Specifically, IATN comprises two attention networks, one of them is to identify the common features between domains through domain classification, and the other aims to extract information from the aspects by using the common features as a bridge. Then, we conduct interactive attention learning for those two networks so that both the sentences and the aspects can influence the final sentiment representation. Extensive experiments on the Amazon reviews dataset and crowdfunding reviews dataset not only demonstrate the effectiveness and universality of our method, but also give an interpretable way to track the attention information for sentiment.

We study the problem of question retrieval in community question answering (CQA). The biggest challenge within this task is lexical gaps between questions since similar questions are usually expressed with different but semantically related words. To bridge the gaps, state-of-the-art methods incorporate extra information such as word-to-word translation and categories of questions into the traditional language models. We find that the existing language model based methods can be interpreted using a new framework, that is they represent words and question categories in a vector space and calculate question-question similarities with a linear combination of dot products of the vectors. The problem is that these methods are either heuristic on data representation or difficult to scale up. We propose a principled and efficient approach to learning representations of data in CQA. In our method, we simultaneously learn vectors of words and vectors of question categories by optimizing an objective function naturally derived from the framework. In question retrieval, we incorporate learnt representations into traditional language models in an effective and efficient way. We conduct experiments on large scale data from Yahoo! Answers and Baidu Knows, and compared our method with state-of-the-art methods on two public data sets. Experimental results show that our method can significantly improve on baseline methods for retrieval relevance. On 1 million training data, our method takes less than 50 minutes to learn a model on a single multicore machine, while the translation based language model needs more than 2 days to learn a translation table on the same machine.

This paper studies the problem of question retrieval in community question answering (CQA). To bridge lexical gaps in questions, which is regarded as the biggest challenge in retrieval, state-of-the-art methods learn translation models using answers under an assumption that they are parallel texts. In practice, however, questions and answers are far from "parallel". Indeed, they are heterogeneous for both the literal level and user behaviors. There are a particularly large number of low quality answers, to which the performance of translation models is vulnerable. To address these problems, we propose a supervised question-answer topic modeling approach. The approach assumes that questions and answers share some common latent topics and are generated in a "question language" and "answer language" respectively following the topics. The topics also determine an answer quality signal. Compared with translation models, our approach not only comprehensively models user behaviors on CQA portals, but also highlights the instinctive heterogeneity of questions and answers. More importantly, it takes answer quality into account and performs robustly against noise in answers. With the topic modeling approach, we propose a topic-based language model, which matches questions not only on a term level but also on a topic level. We conducted experiments on large scale data from Yahoo! Answers and Baidu Knows. Experimental results show that the proposed model can significantly outperform state-of-the-art retrieval models in CQA.