论文检索

Sai Ashish Somayajula, Bokai Hu, Qi Cao, Xin Pan, Pengtao Xie

Instruction-fine-tuned large language models (LLMs) under 14B parameters continue to underperform on natural language understanding (NLU) tasks, often trailing smaller models like BERT-base on benchmarks such as GLUE and SuperGLUE. Motivated by the success of reinforcement learning in reasoning tasks (e.g., DeepSeek), we explore Proximal Policy Optimization (PPO) as a framework to improve the NLU capabilities of LLMs. We frame NLU as a reinforcement learning environment, treating token generation as a sequence of actions and optimizing for reward signals based on alignment with ground-truth labels. PPO consistently outperforms supervised fine-tuning, yielding an average improvement of 6.3 points on GLUE, and surpasses zero-shot and few-shot prompting by 38.7 and 26.1 points, respectively. Notably, PPO-tuned models outperform GPT-4o by over 4% on average across sentiment and natural language inference tasks, including gains of 7.3% on the Mental Health dataset and 10.9% on SIGA-nli. This work highlights a promising direction for adapting LLMs to new tasks by reframing them as reinforcement learning problems, enabling learning through simple end-task rewards rather than extensive data curation. Our code is available at https://github.com/coder-qicao/RL4GLUE.

Jiazhou Ji, Xinru Lu

Large models can achieve higher performance on complex problems through iterative self-reflection. Yet when reflection is uncontrolled, it often leads to longer outputs, higher inference cost, and an increased risk of hallucination. Existing training methods rarely address this trade off. We introduce ReFLAIR, a unified framework that teaches multimodal large models to perform structured reflection via an explicit $think re-think answer $ format and hybrid reward learning. ReFLAIR begins with supervised cold start training on the ReFLAIR-cold dataset of curated multimodal reasoning trajectories, and then trains a Reflection Quality Scorer (RQS) to quantify the utility of rethinking steps. A modified Group Relative Policy Optimization algorithm optimizes a hybrid reward that combines answer correctness, structural fidelity, reflection utility, and sample difficulty. Evaluated on challenging mathematical benchmarks including MathVista, MathVerse, MM-Math and GSM8K, ReFLAIR yields improvements up to +12.2% absolute accuracy, produces higher quality reflective traces, and reduces harmful or redundant revisions. An adaptive test time reflection scheduler further reduces inference cost by nearly 23% while maintaining or improving accuracy. These results demonstrate that structured, reward guided reflection offers a scalable pathway to more reliable and interpretable reasoning in multimodal models.

Henry Hengyuan Zhao, Wenqi Pei, Yifei Tao, Haiyang Mei, Mike Zheng Shou

Existing benchmarks do not test Large Multimodal Models (LMMs) on their interactive intelligence with human users which is vital for developing general-purpose AI assistants. We design InterFeedback, an interactive framework, which can be applied to any LMM and dataset to assess this ability autonomously. On top of this, we introduce InterFeedback-Bench that evaluates interactive intelligence using two representative datasets, MMMU-Pro and MathVerse, to test 10 different open-source LMMs. Additionally, we present InterFeedback-Human, a newly collected dataset of 120 cases designed for manually testing interactive performance in leading models such as OpenAI-o1 and Claude-3.5-Sonnet. Our evaluation results show that state-of-the-art LMM (e.g., OpenAI-o1) can correct their results through human feedback less than 50%. Our findings point to the need for methods that can enhance LMMs’ capabilities to interpret and benefit from feedback.

Zirui Song, Bin Yan, Yuhan Liu, Miao Fang, Mingzhe Li, Rui Yan, Xiuying Chen

Large Language Models (LLMs) have demonstrated remarkable success in various tasks such as natural language understanding, text summarization, and machine translation. However, their general-purpose nature often limits their effectiveness in domain-specific applications that require specialized knowledge, such as healthcare, chemistry, or legal analysis. To address this, researchers have explored diverse methods to enhance LLMs by integrating domain-specific knowledge. In this survey, we provide a comprehensive overview of these methods, which we categorize into four key approaches: dynamic knowledge injection, static knowledge embedding, modular adapters, and prompt optimization. Each approach offers unique mechanisms to equip LLMs with domain expertise, balancing trade-offs between flexibility, scalability, and efficiency. We discuss how these methods enable LLMs to tackle specialized tasks, compare their advantages and disadvantages, evaluate domain-specific LLMs against general LLMs, and highlight the challenges and opportunities in this emerging field. For those interested in delving deeper into this area, we also summarize the commonly used datasets and benchmarks. To keep researchers updated on the latest studies, we maintain an open-source at: blueofficial-repo.com, dedicated to documenting research in the field of specialized LLM.

Wenyu Wang, Mengqi Zhang, Xiaotian Ye, Zhaochun Ren, Pengjie Ren, Zhumin Chen

Large Language Models (LLMs) inevitably acquire harmful information during training on massive datasets. LLM unlearning aims to eliminate the influence of such harmful information while maintaining the model’s overall performance. Existing unlearning methods, represented by gradient ascent-based approaches, primarily focus on forgetting target data while overlooking the crucial impact of logically related knowledge on the effectiveness of unlearning. In this paper, through both theoretical and experimental analyses, we first demonstrate that a key reason for the suboptimal unlearning performance is that models can reconstruct the target content through reasoning with logically related knowledge. To address this issue, we propose Unlearning Improvement via Parameter Extrapolation (UIPE), a method that removes knowledge highly correlated with the forgetting targets. Experimental results show that UIPE significantly enhances the performance of GA-based method and its variants on the TOFU and WMDP benchmarks.

Arti Rani, Shweta Singh, Nihar Ranjan Sahoo, Gaurav Kumar Nayak

Large Language Models (LLMs) have achieved remarkable success on question answering (QA) tasks, yet they often encode harmful biases that compromise fairness and trustworthiness. Most existing bias mitigation approaches are restricted to predefined categories, limiting their ability to address novel or context-specific emergent biases. To bridge this gap, we tackle the novel problem of open-set bias detection and mitigation in text-based QA. We introduce _OpenBiasBench_, a comprehensive benchmark designed to evaluate biases across a wide range of categories and subgroups, encompassing both known and previously unseen biases. Additionally, we propose _Open-DeBias_, a novel, data-efficient, and parameter-efficient debiasing method that leverages adapter modules to mitigate existing social and stereotypical biases while generalizing to unseen ones. Compared to the state-of-the-art BMBI method, Open-DeBias improves QA accuracy on BBQ dataset by nearly **48%** on ambiguous subsets and **6%** on disambiguated ones, using adapters fine-tuned on just a small fraction of the training data. Remarkably, the same adapters, in a zero-shot transfer to Korean BBQ, achieve **84% accuracy**, demonstrating robust language-agnostic generalization. Through extensive evaluation, we also validate the effectiveness of Open-DeBias across a broad range of NLP tasks, including StereoSet and CrowS-Pairs, highlighting its robustness, multilingual strength, and suitability for general-purpose, open-domain bias mitigation. The project page is available at: [https://sites.google.com/view/open-debias25](https://sites.google.com/view/open-debias25)

Pouya Pezeshkpour, Moin Aminnaseri, Estevam Hruschka

Vision-language models (VLMs) have demonstrated impressive performance by effectively integrating visual and textual information to solve complex tasks. However, it is not clear how these models reason over the visual and textual data together, nor how the flow of information between modalities is structured. In this paper, we examine how VLMs reason by analyzing their biases when confronted with scenarios that present conflicting image and text cues—a common occurrence in real-world applications. To uncover the extent and nature of these biases, we build upon existing benchmarks to create five datasets containing mismatched image-text pairs, covering topics in mathematics, science, and visual descriptions. Our analysis shows that VLMs favor text in simpler queries but shift toward images as query complexity increases. This bias correlates with model scale, with the difference between the percentage of image- and text-preferred responses ranging from +56.8% (image favored) to -85.1% (text favored), depending on the task and model. In addition, we explore three mitigation strategies: simple prompt modifications, modifications that explicitly instruct models on how to handle conflicting information (akin to chain-of-thought prompting), and a task decomposition strategy that analyzes each modality separately before combining their results. Our findings indicate that the effectiveness of these strategies in identifying and mitigating bias varies significantly and is closely linked to the model’s overall performance on the task and the specific modality in question. We released our dataset and code.

Parker Seegmiller, Kartik Mehta, Soumya Saha, Chenyang Tao, Shereen Oraby, Arpit Gupta, Tagyoung Chung, Mohit Bansal, Nanyun Peng

Recent works improving LLM math reasoning with synthetic data have used unique setups, making comparison of data synthesis strategies impractical. This leaves many unanswered questions about the roles of different factors in the synthetic data pipeline, such as the impact of filtering low-quality problems. To address this gap, we introduce FLAMES, a Framework for LLM Assessment of Math rEasoning Data Synthesis, and perform a systematic study of 10 existing data synthesis strategies and multiple other factors impacting the performance of synthetic math reasoning data. Our FLAMES experiments provide several valuable insights about the optimal balance of difficulty and diversity of synthetic data. First, data agents designed to increase problem complexity lead to best improvements on most math metrics. Second, with a fixed data generation budget, keeping higher problem coverage is more important than keeping only problems with reliable solutions. Third, GSM8K- and MATH-based synthetic data can lead to improvements on competition-level benchmarks, showcasing easy-to-hard generalization. Leveraging insights from our FLAMES experiments, we design two novel data synthesis strategies for improving out-of-domain generalization and robustness. Further, we develop the FLAMES dataset, an effective blend of our novel and existing data synthesis strategies, outperforming public datasets on OlympiadBench (+15.7), CollegeMath (+4.5), GSMPlus (+6.5), and MATH (+3.1). Fine-tuning Qwen2.5-Math-7B on the FLAMES dataset achieves 81.4% on MATH, surpassing larger Llama3 405B, GPT-4o and Claude 3.5 Sonnet.

Muhammed Saeed, Shaina Raza, Ashmal Vayani, Muhammad Abdul-Mageed, Ali Emami, Shady Shehata

Research on bias in Text-to-Image (T2I) models has primarily focused on demographic representation and stereotypical attributes, overlooking a fundamental question: how does grammatical gender influence visual representation across languages? We introduce a cross-linguistic benchmark examining words where grammatical gender contradicts stereotypical gender associations (e.g., “une sentinelle” - grammatically feminine in French but referring to the stereotypically masculine concept “guard”). Our dataset spans five gendered languages (French, Spanish, German, Italian, Russian) and two gender-neutral control languages (English, Chinese), comprising 800 unique prompts that generated 28,800 images across three state-of-the-art T2I models. Our analysis reveals that grammatical gender dramatically influences image generation: masculine grammatical markers increase male representation to 73% on average (compared to 22% with gender-neutral English), while feminine grammatical markers increase female representation to 38% (compared to 28% in English). These effects vary systematically by language resource availability and model architecture, with high-resource languages showing stronger effects. Our findings establish that language structure itself, not just content, shapes AI-generated visual outputs, introducing a new dimension for understanding bias and fairness in multilingual, multimodal systems.

Woojun Jung, Junyeong Kim

Video-to-text summarization remains underexplored in terms of comprehensive evaluation methods. Traditional n-gram overlap-based metrics and recent large language model (LLM)-based approaches depend heavily on human-written reference summaries, limiting their practicality and sensitivity to nuanced semantic aspects. In this paper, we propose QEVA, a reference-free metric evaluating candidate summaries directly against source videos through multimodal question answering. QEVA assesses summaries along three clear dimensions: Coverage, Factuality, and Temporal Coherence. We also introduce MLVU(VS)-Eval, a new annotated benchmark derived from the MLVU dataset, comprising 800 summaries generated from 200 videos using state-of-the-art video-language multimodal models. This dataset establishes a transparent and consistent framework for evaluation. Experimental results demonstrate that QEVA shows higher correlation with human judgments compared to existing approaches, as measured by Kendall’s \tau_b, \tau_c, and Spearman’s \rho. We hope that our benchmark and metric will facilitate meaningful progress in video-to-text summarization research and provide valuable insights for the development of future evaluation methods.

Yaoshu Wang, Mengyi Yan, Wei Wang

Entity resolution is a fundamental problem in data management that aims to identify all duplicate entries within collections of multi-attribute tuples. Most existing works focus on supervised learning, relying on large amounts of high-quality labeled data, including both positive and negative tuple pairs that are meticulously prepared. However, in reality, the manual annotation process is labor-intensive; in particular, selecting high-quality negative data for labeling is both important and challenging. In this paper, we propose an end-to-end ER solution, PUER, to address low-resource entity resolution (ER) by leveraging Large Language Models (LLMs) in a Positive-Unlabeled (PU) learning setting, where only a small number of positively labeled examples, e.g., 50, and unlabeled data are provided. Unlike directly fine-tuning LLMs in a supervised manner, we solve the entity matching task using reinforcement learning and propose a self-adaptive reward function in the process of RL. To enhance performance, we design an iterative workflow based on the co-training mechanism that fully utilizes entity blocking component to assist the entity matching. This workflow aims to improve the robustness and quality of pseudo-labels so that the performance of entity matching improves. Comprehensive experimental results on various benchmark datasets demonstrate the superiority of PUER. Full version and code are available.

Ahmed Heakl, Sarim Hashmi, Chaimaa Abi, Celine Lee, Abdulrahman Mahmoud

The hardware ecosystem is rapidly evolving, with increasing interest in translating low-level programs across different *instruction set architectures* (ISAs) in a quick, flexible, and correct way to enhance the portability and longevity of existing code. A particularly challenging class of this transpilation problem is translating between complex- (CISC) and reduced- (RISC) hardware architectures, due to fundamental differences in instruction complexity, memory models, and execution paradigms. In this work, we introduce GG (**G**uaranteed **G**uess), an ISA-centric transpilation pipeline that combines the translation power of pre-trained large language models (LLMs) with the rigor of established software testing constructs. Our method generates candidate translations using an LLM from one ISA to another, and embeds such translations within a software-testing framework to build quantifiable confidence in the translation. We evaluate our GG approach over two diverse datasets, enforce high code coverage (>98%) across unit tests, and achieve functional/semantic correctness of 99% on HumanEval programs and 49% on BringupBench programs, respectively. Further, we compare our approach to the state-of-the-art Rosetta 2 framework on Apple Silicon, showcasing 1.73\times faster runtime performance, 1.47\times better energy efficiency, and 2.41\times better memory usage for our transpiled code, demonstrating the effectiveness of GG for real-world CISC-to-RISC translation tasks. We will open-source our codes, data, models, and benchmarks to establish a common foundation for ISA-level code translation research.

Zhihao Zhang, Yiran Zhang, Xiyue Zhou, Liting Huang, Imran Razzak, Preslav Nakov, Usman Naseem

Infodemics and health misinformation have significant negative impact on individuals and society, exacerbating confusion and increasing hesitancy in adopting recommended health measures. Recent advancements in generative AI, capable of producing realistic, human-like text and images, have significantly accelerated the spread and expanded the reach of health misinformation, resulting in an alarming surge in its dissemination. To combat the infodemics, most existing work has focused on developing misinformation datasets from social media and fact-checking platforms, but has faced limitations in topical coverage, inclusion of AI-generation, and accessibility of raw content. To address these gaps, we present MM-Health, a large scale multimodal misinformation dataset in the health domain consisting of 34,746 news article encompassing both textual and visual information. MM-Health includes human-generated multimodal information (5,776 articles) and AI-generated multimodal information (28,880 articles) from various SOTA generative AI models. Additionally, We benchmarked our dataset against three tasks—reliability checks, originality checks, and fine-grained AI detection—demonstrating that existing SOTA models struggle to accurately distinguish the reliability and origin of information. Our dataset aims to support the development of misinformation detection across various health scenarios, facilitating the detection of human and machine-generated content at multimodal levels.

Pranav Bhagat, K N Ajay Shastry, Pranoy Panda, Chaitanya Devaguptapu

Compound AI (CAI) systems, also referred to as LLM Agents, combine LLMs with retrievers and tools to enable information-seeking applications in the real-world. Thus, ensuring these systems perform reliably is critical. However, traditional evaluation using benchmark datasets and aggregate metrics often fails to capture their true operational performance. This is because understanding the operational efficacy of these information-seeking systems requires the ability to probe their behavior across a spectrum of simulated scenarios to identify potential failure modes. Thus, we present a behavior-driven evaluation framework that generates test specifications - explicit descriptions of expected system behaviors in specific scenarios - aligned with real usage contexts. These test specifications serve as formal declarations of system requirements that are then automatically transformed into concrete test cases. Specifically, our framework operates in two phases: (1) generating diverse test specifications via submodular optimization over semantic diversity and document coverage of the tests, and (2) implementing these specifications through graph-based pipelines supporting both tabular and textual sources. Evaluations on QuAC & HybriDialogue datasets, across SoTA LLMs, reveal that our framework identifies failure modes missed by traditional metrics, demonstrating failure rates twice as high as human-curated datasets.

Jian Chen, Zhenyan Chen, Xuming Hu, Peilin Zhou, Yining Hua, Han Fang, Cissy Hing Yee Choy, Xinmei Ke, Jingfeng Luo, Zixuan Yuan

Natural Language to SQL (NL2SQL) provides a new model-centric paradigm that simplifies database access for non-technical users by converting natural language queries into SQL commands. Recent advancements, particularly those integrating Retrieval-Augmented Generation (RAG) and Chain-of-Thought (CoT) reasoning, have made significant strides in enhancing NL2SQL performance. However, challenges such as inaccurate task decomposition and keyword extraction by LLMs remain major bottlenecks, often leading to errors in SQL generation. While existing datasets aim to mitigate these issues by fine-tuning models, they struggle with over-fragmentation of tasks and lack of domain-specific keyword annotations, limiting their effectiveness.To address these limitations, we present DeKeyNLU, a novel dataset which contains 1,500 meticulously annotated QA pairs aimed at refining task decomposition and enhancing keyword extraction precision for the RAG pipeline. Fine-tuned with DeKeyNLU, we propose DeKeySQL, a RAG-based NL2SQL pipeline that employs three distinct modules for user question understanding, entity retrieval, and generation to improve SQL generation accuracy. We benchmarked multiple model configurations within DeKeySQL RAG pipeline. Experimental results demonstrate that fine-tuning with DeKeyNLU significantly improves SQL generation accuracy on both BIRD (62.31% to 69.10%) and Spider (84.2% to 88.7%) dev datasets.

Lei Li, Xiangxu Zhang, Xiao Zhou, Zheng Liu

Medical information retrieval (MIR) is vital for accessing knowledge from electronic health records, scientific literature, and medical databases, supporting applications such as medical education, patient queries, and clinical diagnosis. However, effective zero-shot dense retrieval in the medical domain remains difficult due to the scarcity of relevance-labeled data. To address this challenge, we propose **S**elf-**L**earning **Hy**pothetical **D**ocument **E**mbeddings (**SL-HyDE**), a framework that leverages large language models (LLMs) to generate hypothetical documents conditioned on a query. These documents encapsulate essential medical context, guiding dense retrievers toward the most relevant results. SL-HyDE further employs a self-learning mechanism that iteratively improves pseudo-document generation and retrieval using unlabeled corpora, eliminating the need for labeled data. In addition, we introduce the Chinese Medical Information Retrieval Benchmark (CMIRB), a comprehensive evaluation suite reflecting real-world medical scenarios, comprising five tasks and ten datasets. By benchmarking ten models on CMIRB, we provide a rigorous standard for evaluating MIR systems. Experimental results demonstrate that SL-HyDE significantly outperforms HyDE in retrieval accuracy, while exhibiting strong generalization and scalability across diverse LLM and retriever configurations. Our code and data are publicly available at: https://github.com/ll0ruc/AutoMIR.

Sunkyung Lee, Seongmin Park, Jonghyo Kim, Mincheol Yoon, Jongwuk Lee

Generative recommendation has emerged as a promising paradigm that formulates the recommendations into a text-to-text generation task, harnessing the vast knowledge of large language models. However, existing studies focus on considering the sequential order of items and neglect to handle the temporal dynamics across items, which can imply evolving user preferences. To address this limitation, we propose a novel model, Generative Recommender Using Time awareness (GRUT), effectively capturing hidden user preferences via various temporal signals. We first introduce Time-aware Prompting, consisting of two key contexts. The user-level temporal context models personalized temporal patterns across timestamps and time intervals, while the item-level transition context provides transition patterns across users. We also devise Trend-aware Inference, a training-free method that enhances rankings by incorporating trend information about items with generation likelihood. Extensive experiments demonstrate that GRUT outperforms state-of-the-art models, with gains of up to 15.4% and 14.3% in Recall@5 and NDCG@5 across four benchmark datasets. The source code is available at https://github.com/skleee/GRUT.

Anushka Sivakumar, Andrew Zhang, Zaber Ibn Abdul Hakim, Chris Thomas

This work introduces SteerVLM, a lightweight steering module designed to guide Vision-Language Models (VLMs) towards outputs that better adhere to desired instructions. Our approach learns from the latent embeddings of paired prompts encoding target and converse behaviors to dynamically adjust activations connecting the language modality with image context. This allows for fine-grained, inference-time control over complex output semantics without modifying model weights while preserving performance on off-target tasks. Our steering module requires learning parameters equal to 0.14% of the original VLM’s size. Our steering module gains model control through dimension-wise activation modulation and adaptive steering across layers without requiring pre-extracted static vectors or manual tuning of intervention points. Furthermore, we introduce VNIA (Visual Narrative Intent Alignment), a multimodal dataset specifically created to facilitate the development and evaluation of VLM steering techniques. Our method outperforms existing intervention techniques on steering and hallucination mitigation benchmarks for VLMs and proposes a robust solution for multimodal model control through activation engineering.

Yoel Ashkenazi, Etzion Harari, Regev Yehezkel Imra, Naphtali Abudarham, Dekel Cohen, Yoram Louzoun

Knowledge discovery from large-scale, heterogeneous textual corpora presents a significant challenge. Document clustering offers a practical solution by organizing unstructured texts into coherent groups based on content and thematic similarity. However, clustering does not inherently ensure thematic consistency. Here, we propose a novel framework that constructs a similarity graph over document embeddings and applies iterative graph-based clustering algorithms to partition the corpus into initial clusters. To overcome the limitations of conventional methods in producing semantically consistent clusters, we incorporate iterative feedback from a large language model (LLM) to guide the refinement process. The LLM is used to assess cluster quality and adjust edge weights within the graph, promoting better intra-cluster cohesion and inter-cluster separation. The LLM guidance is based on a set of success Rate metrics that we developed to measure the semantic coherence of clusters. Experimental results on multiple benchmark datasets demonstrate that the iterative process and additional user-supplied a priori edges improve the summaries’ consistency and fluency, highlighting the importance of known connections among the documents. The removal of very rare or very frequent sentences has a mixed effect on the quality scores.Our full code is available here: https://github.com/D2CS-sub/D2CS

Md. Atabuzzaman, Andrew Zhang, Chris Thomas

Large Vision-Language Models (LVLMs) have demonstrated impressive performance on vision-language reasoning tasks. However, their potential for zero-shot fine-grained image classification, a challenging task requiring precise differentiation between visually similar categories, remains underexplored. We present a novel method that transforms zero-shot fine-grained image classification into a visual question-answering framework, leveraging LVLMs’ comprehensive understanding capabilities rather than relying on direct class name generation. We enhance model performance through a novel attention intervention technique. We also address a key limitation in existing datasets by developing more comprehensive and precise class description benchmarks. We validate the effectiveness of our method through extensive experimentation across multiple fine-grained image classification benchmarks. Our proposed method consistently outperforms the current state-of-the-art (SOTA) approach, demonstrating both the effectiveness of our method and the broader potential of LVLMs for zero-shot fine-grained classification tasks. Code and Datasets: https://github.com/Atabuzzaman/Fine-grained-classification