Computation and Language
Beyond Over-Refusal: Scenario-Based Diagnostics and Post-Hoc Mitigation for Exaggerated Refusals in LLMs
Shuzhou Yuan, Ercong Nie, Yinuo Sun et al. (+3 more)
Large
language models (LLMs) frequently produce false refusals, declining benign requests that contain terms resembling unsafe queries. We address this challenge by introducing two comprehensive
benchmarks: the Exaggerated Safety
Benchmark (XSB) for single-turn
prompts, annotated with "Focus" keywords that identify refusal-inducing triggers, and the Multi-turn Scenario-based Exaggerated Safety
Benchmark (MS-XSB), which systematically evaluates refusal calibration in realistic, context-rich dialog settings. Our
benchmarks reveal that exaggerated refusals persist across diverse recent LLMs and are especially pronounced in complex, multi-turn scenarios. To mitigate these failures, we leverage post-hoc explanation methods to identify refusal triggers and deploy three lightweight, model-agnostic approaches, ignore-word instructions,
prompt rephrasing, and
attention steering, at inference time, all without retraining or parameter access. Experiments on four instruction-tuned Llama models demonstrate that these strategies substantially improve compliance on safe
prompts while maintaining robust safety protections. Our findings establish a reproducible framework for diagnosing and mitigating exaggerated refusals, highlighting practical pathways to safer and more helpful LLM deployments.
Read the full abstract ▾
All authors Shuzhou Yuan, Ercong Nie, Yinuo Sun, Chenxuan Zhao, William LaCroix, Michael Färber
Computation and Language
DACIP-RC: Domain Adaptive Continual Instruction Pre-Training via Reading Comprehension on Business Conversations
Elena Khasanova, Harsh Saini, Md Tahmid Rahman Laskar et al. (+3 more)
The rapid advancements in Large
Language Models (LLMs) have enabled their adoption in real-world industrial scenarios for various natural
language processing tasks. However, the high inference cost of large-scale LLMs makes their deployment impractical, necessitating the use of smaller models. Despite their efficiency, smaller LLMs lack robust zero-shot instruction-following capabilities across diverse domains, limiting their adaptability to dynamic user requirements. Traditional fine-tuning approaches exacerbate this issue by inducing catastrophic forgetting, reducing the model's generalization ability for unseen tasks. In this paper, we propose Domain Adaptive Continual Instruction Pre-Training via Reading Comprehension (DACIP-RC), a continual pre-training technique that enhances smaller LLMs' domain adaptability for business conversational tasks. Unlike conventional pre-training approaches that rely on next-
token prediction, DACIP-RC generates diverse task instructions and responses via reading comprehension on conversation transcripts, enabling better instruction generalization. Our empirical evaluations demonstrate that DACIP-RC significantly improves zero-shot generalization across a wide range of business conversational tasks, including meeting summarization, action item generation, and call purpose identification. To the best of our knowledge, this is the first work to apply instruction pre-training on business conversational data, providing insights into how industries can leverage proprietary
datasets for domain adaptation.
Read the full abstract ▾
All authors Elena Khasanova, Harsh Saini, Md Tahmid Rahman Laskar, Xue-Yong Fu, Cheng Chen, Shashi Bhushan TN
Computation and Language
AI Knowledge Assist: An Automated Approach for the Creation of Knowledge Bases for Conversational AI Agents
Md Tahmid Rahman Laskar, Julien Bouvier Tremblay, Xue-Yong Fu et al. (+2 more)
The utilization of conversational AI systems by leveraging
Retrieval Augmented Generation (RAG) techniques to solve customer problems has been on the rise with the rapid progress of Large
Language Models (LLMs). However, the absence of a company-specific dedicated knowledge base is a major barrier to the integration of conversational AI systems in contact centers. To this end, we introduce AI Knowledge Assist, a system that extracts knowledge in the form of question-answer (QA) pairs from historical customer-
agent conversations to automatically build a knowledge base. Fine-tuning a lightweight LLM on internal data demonstrates state-of-the-art performance, outperforming larger closed-source LLMs. More specifically, empirical evaluation on 20 companies demonstrates that the proposed AI Knowledge Assist system that leverages the LLaMA-3.1-8B model eliminates the cold-start gap in contact centers by achieving above 90% accuracy in answering information-seeking questions. This enables immediate deployment of RAG-powered chatbots.
Read the full abstract ▾
All authors Md Tahmid Rahman Laskar, Julien Bouvier Tremblay, Xue-Yong Fu, Cheng Chen, Shashi Bhushan TN
Computation and Language
Mitigating Judgment Preference Bias in Large Language Models through Group-Based Polling
Shuliang Liu, Zhipeng Xu, Zhenghao Liu et al. (+6 more)
Large
Language Models (LLMs) as automatic evaluators, commonly referred to as LLM-as-a-Judge, have also attracted growing
attention. This approach plays a vital role in aligning LLMs with human judgments, providing accurate and reliable assessments. However, LLM-based judgment models often exhibit judgment preference bias during the evaluation phase, tending to favor responses generated by themselves, undermining the reliability of their judgments. This paper introduces the Group-Based Polling Optimization (Genii), an unsupervised multi-
agent collaborative optimization framework that mitigates the inherent judgment preference bias of judgment models. Specifically, Genii integrates various LLM-based judgment models into a multi-
agent system and simulates the interactive client-server polling mechanism to optimize each client
agent unsupervisedly. Our experiments demonstrate that Genii outperforms supervised models trained on annotated judgment data, while requiring no human-labeled annotations. Genii consistently improves performance across different client
agents during the polling, even when weaker models act as server
agents. Further analysis reveals that Genii effectively mitigates judgment preference bias of LLM-based judgment models, demonstrating its effectiveness. All codes are available at https://github.com/NEUIR/Genii.
Read the full abstract ▾
All authors Shuliang Liu, Zhipeng Xu, Zhenghao Liu, Yukun Yan, Minghe Yu, Yu Gu, Chong Chen, Huiyuan Xie, Ge Yu
Computation and Language
Interpreting LLM-as-a-Judge Policies via Verifiable Global Explanations
Jasmina Gajcin, Erik Miehling, Rahul Nair et al. (+3 more)
Using LLMs to evaluate text, that is, LLM-as-a-judge, is increasingly being used at scale to augment or even replace human annotations. As such, it is imperative that we understand the potential biases and risks of doing so. In this work, we propose an approach for extracting high-level concept-based global policies from LLM-as-a-Judge. Our approach consists of two
algorithms: 1) CLoVE (Contrastive Local Verifiable Explanations), which generates verifiable, concept-based, contrastive local explanations and 2) GloVE (Global Verifiable Explanations), which uses iterative clustering, summarization and verification to condense local rules into a global policy. We evaluate GloVE on seven standard benchmarking
datasets for content harm detection. We find that the extracted global policies are highly faithful to decisions of the LLM-as-a-Judge. Additionally, we evaluated the robustness of global policies to text perturbations and
adversarial attacks. Finally, we conducted a user study to evaluate user understanding and satisfaction with global policies.
Read the full abstract ▾
All authors Jasmina Gajcin, Erik Miehling, Rahul Nair, Elizabeth Daly, Radu Marinescu, Seshu Tirupathi
Computation and Language · Cross-listed from cs.AI
Can Risk-taking AI-Assistants suitably represent entities
Ali Mazyaki, Mohammad Naghizadeh, Samaneh Ranjkhah Zonouzaghi et al. (+1 more)
Responsible AI demands systems whose behavioral tendencies can be effectively measured, audited, and adjusted to prevent inadvertently nudging users toward risky decisions or
embedding hidden biases in risk aversion. As
language models (LMs) are increasingly incorporated into AI-driven decision support systems, understanding their risk behaviors is crucial for their responsible deployment. This study investigates the manipulability of risk aversion (MoRA) in LMs, examining their ability to replicate human risk preferences across diverse economic scenarios, with a focus on gender-specific attitudes, uncertainty, role-based decision-making, and the manipulability of risk aversion. The results indicate that while LMs such as DeepSeek Reasoner and Gemini-2.0-flash-lite exhibit some
alignment with human behaviors, notable discrepancies highlight the need to refine bio-centric measures of manipulability. These findings suggest directions for refining AI design to better align human and AI risk preferences and enhance ethical decision-making. The study calls for further advancements in model design to ensure that AI systems more accurately replicate human risk preferences, thereby improving their effectiveness in risk management contexts. This approach could enhance the applicability of AI assistants in managing risk.
Read the full abstract ▾
All authors Ali Mazyaki, Mohammad Naghizadeh, Samaneh Ranjkhah Zonouzaghi, Amirhossein Farshi Sotoudeh
Computation and Language
Evaluating LLM-Generated Legal Explanations for Regulatory Compliance in Social Media Influencer Marketing
Haoyang Gui, Thales Bertaglia, Taylor Annabell et al. (+3 more)
The rise of influencer marketing has blurred boundaries between organic content and sponsored content, making the enforcement of legal rules relating to transparency challenging. Effective regulation requires applying legal knowledge with a clear purpose and reason, yet current detection methods of undisclosed sponsored content generally lack legal grounding or operate as opaque "black boxes". Using 1,143 Instagram posts, we compare gpt-5-nano and gemini-2.5-flash-lite under three prompting strategies with controlled levels of legal knowledge provided. Both models perform strongly in classifying content as sponsored or not (F1 up to 0.93), though performance drops by over 10 points on ambiguous cases. We further develop a taxonomy of
reasoning errors, showing frequent citation omissions (28.57%), unclear references (20.71%), and hidden ads exhibiting the highest miscue rate (28.57%). While adding regulatory text to the
prompt improves explanation quality, it does not consistently improve detection accuracy. The contribution of this paper is threefold. First, it makes a novel addition to regulatory compliance technology by providing a taxonomy of common errors in LLM-generated legal
reasoning to evaluate whether automated moderation is not only accurate but also legally robust, thereby advancing the transparent detection of influencer marketing content. Second, it features an original
dataset of LLM explanations annotated by two students who were trained in influencer marketing law. Third, it combines quantitative and qualitative evaluation strategies for LLM explanations and critically reflects on how these findings can support advertising regulatory bodies in automating moderation processes on a solid legal foundation.
Read the full abstract ▾
All authors Haoyang Gui, Thales Bertaglia, Taylor Annabell, Catalina Goanta, Tjomme Dooper, Gerasimos Spanakis
Computation and Language · Cross-listed from cs.IR
VersionRAG: Version-Aware Retrieval-Augmented Generation for Evolving Documents
Daniel Huwiler, Kurt Stockinger and Jonathan Fürst
Retrieval-Augmented Generation (RAG) systems fail when documents evolve through versioning-a ubiquitous characteristic of technical documentation. Existing approaches achieve only 58-64% accuracy on version-sensitive questions, retrieving semantically similar content without temporal validity checks. We present VersionRAG, a version-aware RAG framework that explicitly models document
evolution through a hierarchical graph structure capturing version sequences, content boundaries, and changes between document states. During
retrieval, VersionRAG routes queries through specialized paths based on intent classification, enabling precise version-aware filtering and change tracking. On our VersionQA
benchmark-100 manually curated questions across 34 versioned technical documents-VersionRAG achieves 90% accuracy, outperforming naive RAG (58%) and GraphRAG (64%). VersionRAG reaches 60% accuracy on implicit change detection where baselines fail (0-10%), demonstrating its ability to track undocumented modifications. Additionally, VersionRAG requires 97% fewer
tokens during indexing than GraphRAG, making it practical for large-scale deployment. Our work establishes versioned document QA as a distinct task and provides both a solution and
benchmark for future research.
Read the full abstract ▾
Computation and Language
Lossless Vocabulary Reduction for Auto-Regressive Language Models
Daiki Chijiwa, Taku Hasegawa, Kyosuke Nishida et al. (+4 more)
Tokenization – the process of decomposing a given text into a sequence of subwords called
tokens – is one of the key components in the development of
language models. Particularly, auto-regressive
language models generate texts
token by
token, i.e., by predicting the next-
token distribution given the previous ones, and thus tokenization directly affects their efficiency in text generation. Since each
language model has their own vocabulary as a set of possible
tokens, they struggle to cooperate with each other at the level of next-
token distributions such as model ensemble. In this paper, we establish a theoretical framework of lossless vocabulary reduction, which efficiently converts a given auto-regressive
language model into the one with an arbitrarily small vocabulary without any loss in accuracy. This framework allows
language models with different tokenization to cooperate with each other efficiently by reduction to their maximal common vocabulary. Specifically, we empirically demonstrate its applicability to model ensemble with different tokenization.
Read the full abstract ▾
All authors Daiki Chijiwa, Taku Hasegawa, Kyosuke Nishida, Shin'ya Yamaguchi, Tomoya Ohba, Tamao Sakao, Susumu Takeuchi
Computation and Language
The Price of Thought: A Multilingual Analysis of Reasoning, Performance, and Cost of Negotiation in Large Language Models
Sherzod Hakimov, Roland Bernard, Tim Leiber et al. (+5 more)
Negotiation is a fundamental challenge for AI
agents, as it requires an ability to reason strategically, model opponents, and balance cooperation with competition. We present the first comprehensive study that systematically evaluates how explicit
reasoning training affects the negotiation abilities of both commercial and open-weight large
language models, comparing these models to their vanilla counterparts across three
languages. Using a self-play setup across three diverse dialogue games, we analyse trade-offs between performance and cost, the
language consistency of
reasoning processes, and the nature of strategic adaptation exhibited by models. Our findings show that enabling
reasoning – that is, scaling test time compute – significantly improves negotiation outcomes by enhancing collaboration and helping models overcome task complexities, but comes at a substantial computational cost:
reasoning improves GPT-5's performance by 31.4 % while increasing its cost by nearly 400 %. Most critically, we uncover a significant
multilingual reasoning distinction: open-weight models consistently switch to English for their internal
reasoning steps, even when negotiating in German or Italian (and thus possibly impacting potential explainability gains through the disclosure of
reasoning traces), while a leading commercial model maintains
language consistency between
reasoning and final output.
Read the full abstract ▾
All authors Sherzod Hakimov, Roland Bernard, Tim Leiber, Karl Osswald, Kristina Richert, Ruilin Yang, Raffaella Bernardi, David Schlangen
Computation and Language
Arguments that Alter Minds: LLM Rationales Sway Human (and LLM) Notions of Plausibility
Shramay Palta, Peter Rankel, Sarah Wiegreffe et al. (+1 more)
We investigate the degree to which human (and LLM) plausibility judgments of multiple-choice commonsense
benchmark answers are subject to influence by (im)plausibility arguments for or against an answer, in particular, using rationales generated by LLMs. We collect 3,000 plausibility judgments from humans and another 13,600 judgments from LLMs. Overall, we observe increases and decreases in mean human plausibility ratings in the presence of LLM-generated PRO and CON rationales, respectively, suggesting that, on the whole, human judges find these rationales convincing. Experiments with LLMs reveal similar patterns of influence. Our findings demonstrate a novel use of LLMs for studying aspects of human cognition, while also raising practical concerns that, even in domains where humans are “experts” (i.e., common sense), LLMs have the potential to exert considerable influence on people's beliefs.
Read the full abstract ▾
All authors Shramay Palta, Peter Rankel, Sarah Wiegreffe, Rachel Rudinger
Computation and Language · Cross-listed from cs.AI
AutoQual: An LLM Agent for Automated Discovery of Interpretable Features for Review Quality Assessment
Xiaochong Lan, Jie Feng, Yinxing Liu et al. (+2 more)
Ranking online reviews by their intrinsic quality is a critical task for e-commerce platforms and information services, impacting user experience and business outcomes. However, quality is a domain-dependent and dynamic concept, making its assessment a formidable challenge. Traditional methods relying on hand-crafted features are unscalable across domains and fail to adapt to evolving content patterns, while modern deep learning approaches often produce black-box models that lack interpretability and may prioritize semantics over quality. To address these challenges, we propose AutoQual, an LLM-based
agent framework that automates the discovery of interpretable features. While demonstrated on review quality assessment, AutoQual is designed as a general framework for transforming tacit knowledge embedded in data into explicit, computable features. It mimics a human research process, iteratively generating feature hypotheses through reflection, operationalizing them via autonomous tool implementation, and accumulating experience in a persistent
memory. We deploy our method on a large-scale online platform with a billion-level user base. Large-scale A/B testing confirms its effectiveness, increasing average reviews viewed per user by 0.79% and the conversion rate of review readers by 0.27%.
Read the full abstract ▾
All authors Xiaochong Lan, Jie Feng, Yinxing Liu, Xinlei Shi, Yong Li
Computation and Language
FedDTRE: Federated Dialogue Generation Models Powered by Trustworthiness Evaluation
Shule Lu, Lingxiang Wang, Sijia Wen et al. (+2 more)
With the rapid development of artificial intelligence, dialogue systems have become a prominent form of human-computer interaction. However, traditional centralized or fully local training approaches face challenges in balancing
privacy preservation and personalization due to data
privacy concerns and heterogeneous device capabilities.
Federated learning, as a representative distributed paradigm, offers a promising solution. However, existing methods often suffer from overfitting under limited client data and tend to forget global information after multiple training rounds, leading to poor generalization. To address these issues, we propose FedDTRE, a
Federated adaptive aggregation strategy for Dialogue generation based on Trustworthiness Evaluation. Instead of directly replacing local models with the global model, FedDTRE leverages trustworthiness scores of both global and local models on a
fairness-oriented evaluation
dataset to dynamically regulate the global model's contribution during local updates. Experimental results demonstrate that FedDTRE can improve dialogue model performance and enhance the quality of dialogue generation.
Read the full abstract ▾
All authors Shule Lu, Lingxiang Wang, Sijia Wen, Ziwei Wang, Hainan Zhang
Computation and Language
A Survey of Process Reward Models: From Outcome Signals to Process Supervisions for Large Language Models
Congmin Zheng, Jiachen Zhu, Zhuoying Ou et al. (+8 more)
Although Large
Language Models (LLMs) exhibit advanced
reasoning ability, conventional
alignment remains largely dominated by outcome reward models (ORMs) that judge only final answers. Process Reward Models(PRMs) address this gap by evaluating and guiding
reasoning at the step or trajectory level. This survey provides a systematic overview of PRMs through the full loop: how to generate process data, build PRMs, and use PRMs for test-time scaling and reinforcement learning. We summarize applications across math, code, text,
multimodal reasoning, robotics, and
agents, and review emerging
benchmarks. Our goal is to clarify design spaces, reveal open challenges, and guide future research toward fine-grained, robust
reasoning alignment.
Read the full abstract ▾
All authors Congmin Zheng, Jiachen Zhu, Zhuoying Ou, Yuxiang Chen, Kangning Zhang, Rong Shan, Zeyu Zheng, Mengyue Yang, Jianghao Lin, Yong Yu, Weinan Zhang
Computation and Language · Cross-listed from cs.IR
TaoSR-AGRL: Adaptive Guided Reinforcement Learning Framework for E-commerce Search Relevance
Jianhui Yang, Yiming Jin, Pengkun Jiao et al. (+6 more)
Query-product relevance prediction is fundamental to e-commerce search and has become even more critical in the era of AI-powered shopping, where semantic understanding and complex
reasoning directly shape the user experience and business conversion. Large
Language Models (LLMs) enable generative,
reasoning-based approaches, typically aligned via supervised fine-tuning (SFT) or preference optimization methods like Direct Preference Optimization (DPO). However, the increasing complexity of business rules and user queries exposes the inability of existing methods to endow models with robust
reasoning capacity for long-tail and challenging cases. Efforts to address this via reinforcement learning strategies like Group Relative Policy Optimization (GRPO) often suffer from sparse terminal rewards, offering insufficient guidance for multi-step
reasoning and slowing convergence. To address these challenges, we propose TaoSR-AGRL, an Adaptive Guided Reinforcement Learning framework for LLM-based relevance prediction in Taobao Search Relevance. TaoSR-AGRL introduces two key innovations: (1) Rule-aware Reward Shaping, which decomposes the final relevance judgment into dense, structured rewards aligned with domain-specific relevance criteria; and (2) Adaptive Guided Replay, which identifies low-accuracy rollouts during training and injects targeted ground-truth guidance to steer the policy away from stagnant, rule-violating
reasoning patterns toward compliant trajectories. TaoSR-AGRL was evaluated on large-scale real-world
datasets and through online side-by-side human evaluations on Taobao Search. It consistently outperforms DPO and standard GRPO baselines in offline experiments, improving relevance accuracy, rule adherence, and training stability. The model trained with TaoSR-AGRL has been successfully deployed in the main search scenario on Taobao, serving hundreds of millions of users.
Read the full abstract ▾
All authors Jianhui Yang, Yiming Jin, Pengkun Jiao, Chenhe Dong, Zerui Huang, Shaowei Yao, Xiaojiang Zhou, Dan Ou, Haihong Tang
Computation and Language · Cross-listed from eess.AS
Pseudo2Real: Task Arithmetic for Pseudo-Label Correction in Automatic Speech Recognition
Yi-Cheng Lin, Yu-Hsuan Li Liang, Hsuan Su et al. (+4 more)
Robust ASR under domain shift is crucial because real-world systems encounter unseen accents and domains with limited labeled data. Although pseudo-labeling offers a practical workaround, it often introduces systematic, accent-specific errors that filtering fails to fix. We ask: How can we correct these recurring biases without target ground truth? We propose a simple parameter-space correction: in a source domain containing both real and pseudo-labeled data, two ASR models are fine-tuned from the same initialization, one on ground-truth labels and the other on pseudo-labels, and their weight difference forms a correction vector that captures pseudo-label biases. When applied to a pseudo-labeled target model, this vector enhances recognition, achieving up to a 35% relative Word Error Rate (WER) reduction on AfriSpeech-200 across ten African accents with the Whisper tiny model.
Read the full abstract ▾
All authors Yi-Cheng Lin, Yu-Hsuan Li Liang, Hsuan Su, Tzu-Quan Lin, Shang-Tse Chen, Yun-Nung Chen, Hung-yi Lee
Computation and Language
Ivan Kuznetsov, Jacopo Grassi, Dmitrii Pantiukhin et al. (+3 more)
Large
language models (LLMs) are increasingly deployed for
climate-related applications, where understanding internal climatological knowledge is crucial for reliability and misinformation risk assessment. Despite growing adoption, the capacity of LLMs to recall
climate normals from parametric knowledge remains largely uncharacterized. We investigate the capacity of contemporary LLMs to recall
climate normals without external
retrieval, focusing on a prototypical query: mean July 2-m air temperature 1991-2020 at specified locations. We construct a global grid of queries at 1{\deg} resolution land points, providing coordinates and location descriptors, and validate responses against ERA5 reanalysis. Results show that LLMs encode non-trivial
climate structure, capturing latitudinal and topographic patterns, with root-mean-square errors of 3-6 {\deg}C and biases of $\pm$1 {\deg}C. However, spatially coherent errors remain, particularly in mountains and high latitudes. Performance degrades sharply above 1500 m, where RMSE reaches 5-13 {\deg}C compared to 2-4 {\deg}C at lower elevations. We find that including geographic context (country, city, region) reduces errors by 27% on average, with larger models being most sensitive to location descriptors. While models capture the global mean magnitude of observed warming between 1950-1974 and 2000-2024, they fail to reproduce spatial patterns of temperature change, which directly relate to assessing
climate change. This limitation highlights that while LLMs may capture present-day
climate distributions, they struggle to represent the regional and local expression of long-term shifts in temperature essential for understanding
climate dynamics. Our evaluation framework provides a reproducible
benchmark for quantifying parametric
climate knowledge in LLMs and complements existing
climate communication assessments.
Read the full abstract ▾
All authors Ivan Kuznetsov, Jacopo Grassi, Dmitrii Pantiukhin, Boris Shapkin, Thomas Jung, Nikolay Koldunov
Computation and Language
ChatGPT as a Translation Engine: A Case Study on Japanese-English
Vincent Michael Sutanto, Giovanni Gatti De Giacomo, Toshiaki Nakazawa et al. (+1 more)
This study investigates ChatGPT for Japanese-English translation, exploring simple and enhanced
prompts and comparing against commercially available translation engines. Performing both automatic and MQM-based human evaluations, we found that document-level translation outperforms sentence-level translation for ChatGPT. On the other hand, we were not able to determine if enhanced
prompts performed better than simple
prompts in our experiments. We also discovered that ChatGPT-3.5 was preferred by automatic evaluation, but a tradeoff exists between accuracy (ChatGPT-3.5) and fluency (ChatGPT-4). Lastly, ChatGPT yields competitive results against two widely-known translation systems.
Read the full abstract ▾
All authors Vincent Michael Sutanto, Giovanni Gatti De Giacomo, Toshiaki Nakazawa, Masaru Yamada
Computation and Language
Learning on the Job: An Experience-Driven Self-Evolving Agent for Long-Horizon Tasks
Cheng Yang, Xuemeng Yang, Licheng Wen et al. (+9 more)
Large
Language Models have demonstrated remarkable capabilities across diverse domains, yet significant challenges persist when deploying them as AI
agents for real-world long-horizon tasks. Existing LLM
agents suffer from a critical limitation: they are test-time static and cannot learn from experience, lacking the ability to accumulate knowledge and continuously improve on the job. To address this challenge, we propose MUSE, a novel
agent framework that introduces an experience-driven, self-evolving system centered around a hierarchical
Memory Module. MUSE organizes diverse levels of experience and leverages them to plan and execute long-horizon tasks across multiple applications. After each sub-task execution, the
agent autonomously reflects on its trajectory, converting the raw trajectory into structured experience and integrating it back into the
Memory Module. This mechanism enables the
agent to evolve beyond its static pretrained parameters, fostering continuous learning and self-
evolution. We evaluate MUSE on the long-horizon productivity
benchmark TAC. It achieves new SOTA performance by a significant margin using only a lightweight Gemini-2.5 Flash model. Sufficient Experiments demonstrate that as the
agent autonomously accumulates experience, it exhibits increasingly superior task completion capabilities, as well as robust continuous learning and self-
evolution capabilities. Moreover, the accumulated experience from MUSE exhibits strong generalization properties, enabling zero-shot improvement on new tasks. MUSE establishes a new paradigm for AI
agents capable of real-world productivity task automation.
Read the full abstract ▾
All authors Cheng Yang, Xuemeng Yang, Licheng Wen, Daocheng Fu, Jianbiao Mei, Rong Wu, Pinlong Cai, Yufan Shen, Nianchen Deng, Botian Shi, Yu Qiao, Haifeng Li
Computation and Language
Leveraging Author-Specific Context for Scientific Figure Caption Generation: 3rd SciCap Challenge
Watcharapong Timklaypachara, Monrada Chiewhawan, Nopporn Lekuthai et al. (+1 more)
Scientific figure captions require both accuracy and stylistic consistency to convey visual information. Here, we present a domain-specific caption generation system for the 3rd SciCap Challenge that integrates figure-related textual context with author-specific writing styles using the LaMP-Cap
dataset. Our approach uses a two-stage pipeline: Stage 1 combines context filtering, category-specific
prompt optimization via DSPy's MIPROv2 and SIMBA, and caption candidate selection; Stage 2 applies few-shot prompting with profile figures for stylistic refinement. Our experiments demonstrate that category-specific
prompts outperform both zero-shot and general optimized approaches, improving ROUGE-1 recall by +8.3% while limiting precision loss to -2.8% and BLEU-4 reduction to -10.9%. Profile-informed stylistic refinement yields 40–48% gains in BLEU scores and 25–27% in ROUGE. Overall, our system demonstrates that combining contextual understanding with author-specific stylistic adaptation can generate captions that are both scientifically accurate and stylistically faithful to the source paper.
Read the full abstract ▾
All authors Watcharapong Timklaypachara, Monrada Chiewhawan, Nopporn Lekuthai, Titipat Achakulvisut
Computation and Language · Cross-listed from cs.AI
VoiceAgentBench: Are Voice Assistants ready for agentic tasks?
Dhruv Jain, Harshit Shukla, Gautam Rajeev et al. (+3 more)
Large scale
Speech Language Models have enabled voice assistants capable of understanding natural spoken queries and performing complex tasks. However, existing
speech benchmarks largely focus on isolated capabilities such as transcription or question answering and do not systematically evaluate agentic behavior or
adversarial robustness. To address this, we introduce VoiceAgentBench, a comprehensive
benchmark for evaluating SpeechLMs in realistic spoken agentic settings, comprising 6,000+ synthetic spoken queries spanning single-tool invocations, multi-tool workflows, multi-turn dialogue, and safety evaluations across English and six Indic
languages. To ensure speaker diversity, we further simulate speaker variability using a novel sampling strategy that selects audios for TTS voice conversion based on speaker
embeddings to maximize acoustic diversity. Our evaluation measures tool selection accuracy, structural consistency, and the correctness of tool invocations, including
adversarial robustness. Across agentic tasks, ASR-LLM pipelines outperform end-to-end SpeechLMs, achieving up to 60.6% average parameter-filling accuracy on English, while SpeechLMs exhibit lower performance and sharper degradation on Indic
languages. All models struggle in sequential workflows and safety evaluations, highlighting persistent limitations in tool orchestration,
multilingual generalization, and safety robustness. VoiceAgentBench is publicly available on Hugging Face at https://huggingface.co/
datasets/krutrim-ai-labs/VoiceAgentBench, and the codebase is released at https://github.com/ola-krutrim/VoiceAgentBench.
Read the full abstract ▾
All authors Dhruv Jain, Harshit Shukla, Gautam Rajeev, Ashish Kulkarni, Chandra Khatri, Shubham Agarwal
Computation and Language
Active Confusion Expression in Large Language Models: Leveraging World Models toward Better Social Reasoning
Jialu Du, Guiyang Hou, Yihui Fu et al. (+4 more)
While large
language models (LLMs) excel in mathematical and code
reasoning, we observe they struggle with social
reasoning tasks, exhibiting cognitive confusion, logical inconsistencies, and conflation between objective world states and subjective belief states. Through deteiled analysis of DeepSeek-R1's
reasoning trajectories, we find that LLMs frequently encounter
reasoning impasses and tend to output contradictory terms like "tricky" and "confused" when processing scenarios with multiple participants and timelines, leading to erroneous
reasoning or infinite loops. The core issue is their inability to disentangle objective reality from
agents' subjective beliefs. To address this, we propose an adaptive world model-enhanced
reasoning mechanism that constructs a dynamic textual world model to track entity states and temporal sequences. It dynamically monitors
reasoning trajectories for confusion indicators and promptly intervenes by providing clear world state descriptions, helping models navigate through cognitive dilemmas. The mechanism mimics how humans use implicit world models to distinguish between external events and internal beliefs. Evaluations on three social
benchmarks demonstrate significant improvements in accuracy (e.g., +10% in Hi-ToM) while reducing computational costs (up to 33.8%
token reduction), offering a simple yet effective solution for deploying LLMs in social contexts.
Read the full abstract ▾
All authors Jialu Du, Guiyang Hou, Yihui Fu, Chen Wu, Wenqi Zhang, Yongliang Shen, Weiming Lu
Computation and Language
LightReasoner: Can Small Language Models Teach Large Language Models Reasoning?
Jingyuan Wang, Yankai Chen, Zhonghang Li et al. (+1 more)
Large
language models (LLMs) have demonstrated remarkable progress in
reasoning, often through supervised fine-tuning (SFT). However, SFT is resource-intensive, relying on large curated
datasets, rejection-sampled demonstrations, and uniform optimization across all
tokens, even though only a fraction carry meaningful learning value. In this work, we explore a counterintuitive idea: can smaller
language models (SLMs) teach larger
language models (LLMs) by revealing high-value
reasoning moments that reflect the latter's unique strength? We propose LightReasoner, a novel framework that leverages the behavioral divergence between a stronger expert model (LLM) and a weaker amateur model (SLM). LightReasoner operates in two stages: (1) a sampling stage that pinpoints critical
reasoning moments and constructs supervision examples capturing the expert's advantage through expert-amateur contrast, and (2) a fine-tuning stage that aligns the expert model with these distilled examples, amplifying its
reasoning strengths. Across seven mathematical
benchmarks, LightReasoner improves accuracy by up to 28.1%, while reducing time consumption by 90%, sampled problems by 80%, and tuned
token usage by 99%, all without relying on ground-truth labels. By turning weaker SLMs into effective teaching signals, LightReasoner offers a scalable and resource-efficient approach for advancing LLM
reasoning. Code is available at: https://github.com/HKUDS/LightReasoner
Read the full abstract ▾
All authors Jingyuan Wang, Yankai Chen, Zhonghang Li, Chao Huang
Computation and Language
A$^2$Search: Ambiguity-Aware Question Answering with Reinforcement Learning
Fengji Zhang, Xinyao Niu, Chengyang Ying et al. (+7 more)
Recent advances in Large
Language Models (LLMs) and Reinforcement Learning (RL) have led to strong performance in open-domain question answering (QA). However, existing models still struggle with questions that admit multiple valid answers. Standard QA
benchmarks, which typically assume a single gold answer, overlook this reality and thus produce inappropriate training signals. Existing attempts to handle ambiguity often rely on costly manual annotation, which is difficult to scale to multi-hop
datasets such as HotpotQA and MuSiQue. In this paper, we present A$^2$Search, an annotation-free, end-to-end training framework to recognize and handle ambiguity. At its core is an automated pipeline that detects ambiguous questions and gathers alternative answers via trajectory sampling and evidence verification. The model is then optimized with RL using a carefully designed $\mathrm{AnsF1}$ reward, which naturally accommodates multiple answers. Experiments on eight open-domain QA
benchmarks demonstrate that A$^2$Search achieves new state-of-the-art performance. With only a single rollout, A$^2$Search-7B yields an average $\mathrm{AnsF1}@1$ score of $48.4\%$ across four multi-hop
benchmarks, outperforming all strong baselines, including the substantially larger ReSearch-32B ($46.2\%$). Extensive analyses further show that A$^2$Search resolves ambiguity and generalizes across
benchmarks, highlighting that embracing ambiguity is essential for building more reliable QA systems. Our code, data, and model weights can be found at https://github.com/zfj1998/A2Search
Read the full abstract ▾
All authors Fengji Zhang, Xinyao Niu, Chengyang Ying, Guancheng Lin, Zhongkai Hao, Zhou Fan, Chengen Huang, Jacky Keung, Bei Chen, Junyang Lin
Computation and Language · Cross-listed from cs.CV
TTOM: Test-Time Optimization and Memorization for Compositional Video Generation
Leigang Qu, Ziyang Wang, Na Zheng et al. (+3 more)
Video Foundation Models (VFMs) exhibit remarkable visual generation performance, but struggle in compositional scenarios (e.g., motion, numeracy, and spatial relation). In this work, we introduce Test-Time Optimization and Memorization (TTOM), a training-free framework that aligns VFM outputs with spatiotemporal layouts during inference for better text-image
alignment. Rather than direct intervention to latents or
attention per-sample in existing work, we integrate and optimize new parameters guided by a general layout-
attention objective. Furthermore, we formulate
video generation within a streaming setting, and maintain historical optimization contexts with a parametric
memory mechanism that supports flexible operations, such as insert, read, update, and delete. Notably, we found that TTOM disentangles compositional world knowledge, showing powerful transferability and generalization. Experimental results on the T2V-CompBench and Vbench
benchmarks establish TTOM as an effective, practical, scalable, and efficient framework to achieve cross-modal
alignment for compositional
video generation on the fly.
Read the full abstract ▾
All authors Leigang Qu, Ziyang Wang, Na Zheng, Wenjie Wang, Liqiang Nie, Tat-Seng Chua
Computation and Language
Vision-Enabled LLMs in Historical Lexicography: Digitising and Enriching Estonian-German Dictionaries from the 17th and 18th Centuries
Madis Jürviste and Joonatan Jakobson
This article presents research conducted at the Institute of the Estonian
Language between 2022 and 2025 on the application of large
language models (LLMs) to the study of 17th and 18th century Estonian dictionaries. The authors address three main areas: enriching historical dictionaries with modern word forms and meanings; using
vision-enabled LLMs to perform text recognition on sources printed in Gothic script (Fraktur); and preparing for the creation of a unified, cross-source
dataset. Initial experiments with J. Gutslaff's 1648 dictionary indicate that LLMs have significant potential for semi-automatic enrichment of dictionary information. When provided with sufficient context, Claude 3.7 Sonnet accurately provided meanings and modern equivalents for 81% of headword entries. In a text recognition experiment with A. T. Helle's 1732 dictionary, a zero-shot method successfully identified and structured 41% of headword entries into error-free JSON-formatted output. For digitising the Estonian-German dictionary section of A. W. Hupel's 1780 grammar, overlapping tiling of scanned image files is employed, with one LLM being used for text recognition and a second for merging the structured output. These findings demonstrate that even for minor
languages LLMs have a significant potential for saving time and financial resources.
Read the full abstract ▾
Computation and Language
Comprehensiveness Metrics for Automatic Evaluation of Factual Recall in Text Generation
Adam Dejl, James Barry, Alessandra Pascale et al. (+1 more)
Despite demonstrating remarkable performance across a wide range of tasks, large
language models (LLMs) have also been found to frequently produce outputs that are incomplete or selectively omit key information. In sensitive domains, such omissions can result in significant harm comparable to that posed by factual inaccuracies, including
hallucinations. In this study, we address the challenge of evaluating the comprehensiveness of LLM-generated texts, focusing on the detection of missing information or underrepresented viewpoints. We investigate three automated evaluation metrics: (1) an NLI-based method that decomposes texts into atomic statements and uses natural
language inference (NLI) to identify missing facts, (2) a Q&A-based metric that extracts question-answer pairs and compares responses across sources, and (3) an end-to-end approach that directly identifies missing content using LLMs. Our experiments demonstrate the surprising effectiveness of the simple end-to-end metric compared to more complex metrics, though at the cost of reduced robustness, interpretability and result granularity. We further assess the comprehensiveness of responses from several popular open-weight LLMs when answering user queries based on multiple sources.
Read the full abstract ▾
All authors Adam Dejl, James Barry, Alessandra Pascale, Javier Carnerero Cano
Computation and Language
STEPER: Step-wise Knowledge Distillation for Enhancing Reasoning Ability in Multi-Step Retrieval-Augmented Language Models
Kyumin Lee, Minjin Jeon, Sanghwan Jang et al. (+1 more)
Answering complex real-world questions requires step-by-step
retrieval and integration of relevant information to generate well-grounded responses. However, existing knowledge
distillation methods overlook the need for different
reasoning abilities at different steps, hindering transfer in multi-step
retrieval-augmented frameworks. To address this, we propose Stepwise Knowledge
Distillation for Enhancing
Reasoning Ability in Multi-Step
Retrieval-Augmented
Language Models (StepER). StepER employs step-wise supervision to align with evolving information and
reasoning demands across stages. Additionally, it incorporates difficulty-aware training to progressively optimize learning by prioritizing suitable steps. Our method is adaptable to various multi-step
retrieval-augmented
language models, including those that use
retrieval queries for
reasoning paths or decomposed questions. Extensive experiments show that StepER outperforms prior methods on multi-hop QA
benchmarks, with an 8B model achieving performance comparable to a 70B teacher model.
Read the full abstract ▾
All authors Kyumin Lee, Minjin Jeon, Sanghwan Jang, Hwanjo Yu
Computation and Language
Towards Human-Like Grading: A Unified LLM-Enhanced Framework for Subjective Question Evaluation
Fanwei Zhua, Jiaxuan He, Xiaoxiao Chen et al. (+3 more)
Automatic grading of subjective questions remains a significant challenge in examination assessment due to the diversity in question formats and the open-ended nature of student responses. Existing works primarily focus on a specific type of subjective question and lack the generality to support comprehensive exams that contain diverse question types. In this paper, we propose a unified Large
Language Model (LLM)-enhanced auto-grading framework that provides human-like evaluation for all types of subjective questions across various domains. Our framework integrates four complementary modules to holistically evaluate student answers. In addition to a basic text matching module that provides a foundational assessment of content similarity, we leverage the powerful
reasoning and generative capabilities of LLMs to: (1) compare key knowledge points extracted from both student and reference answers, (2) generate a pseudo-question from the student answer to assess its relevance to the original question, and (3) simulate human evaluation by identifying content-related and non-content strengths and weaknesses. Extensive experiments on both general-purpose and domain-specific
datasets show that our framework consistently outperforms traditional and LLM-based baselines across multiple grading metrics. Moreover, the proposed system has been successfully deployed in real-world training and certification exams at a major e-commerce enterprise.
Read the full abstract ▾
All authors Fanwei Zhua, Jiaxuan He, Xiaoxiao Chen, Zulong Chen, Quan Lu, Chenrui Mei
Computation and Language
ACE: Attribution-Controlled Knowledge Editing for Multi-hop Factual Recall
Jiayu Yang, Yuxuan Fan, Songning Lai et al. (+5 more)
Large
Language Models (LLMs) require efficient knowledge editing (KE) to update factual information, yet existing methods exhibit significant performance decay in multi-hop factual recall. This failure is particularly acute when edits involve intermediate implicit subjects within
reasoning chains. Through causal analysis, we reveal that this limitation stems from an oversight of how chained knowledge is dynamically represented and utilized at the
neuron level. We discover that during multi hop
reasoning, implicit subjects function as query
neurons, which sequentially activate corresponding value
neurons across
transformer layers to accumulate information toward the final answer, a dynamic prior KE work has overlooked. Guided by this insight, we propose ACE: Attribution-Controlled Knowledge Editing for Multi-hop Factual Recall, a framework that leverages
neuron-level attribution to identify and edit these critical query-value (Q-V) pathways. ACE provides a mechanistically grounded solution for multi-hop KE, empirically outperforming state-of-the-art methods by 9.44% on GPT-J and 37.46% on Qwen3-8B. Our analysis further reveals more fine-grained activation patterns in Qwen3 and demonstrates that the semantic interpretability of value
neurons is orchestrated by query-driven accumulation. These findings establish a new pathway for advancing KE capabilities based on the principled understanding of internal
reasoning mechanisms.
Read the full abstract ▾
All authors Jiayu Yang, Yuxuan Fan, Songning Lai, Shengen Wu, Jiaqi Tang, Chun Kang, Zhijiang Guo, Yutao Yue