Peer Revenue Fully funded

Papers

Fresh from arXiv, typeset properly. Highlighted words trade on the market.

Computer Science Computation and Language

cs.CL Watch Computation and Language

Tuesday 2 December

Computation and Language

Towards Unification of Hallucination Detection and Fact Verification for Large Language Models

Weihang Su, Jianming Long, Changyue Wang et al. (+5 more)

Large Language Models (LLMs) frequently exhibit hallucinations, generating content that appears fluent and coherent but is factually incorrect. Such errors undermine trust and hinder their adoption in real-world applications. To address this challenge, two distinct research paradigms have emerged: model-centric Hallucination Detection (HD) and text-centric Fact Verification (FV). Despite sharing the same goal, these paradigms have evolved in isolation, using distinct assumptions, datasets, and evaluation protocols. This separation has created a research schism that hinders their collective progress. In this work, we take a decisive step toward bridging this divide. We introduce UniFact, a unified evaluation framework that enables direct, instance-level comparison between FV and HD by dynamically generating model outputs and corresponding factuality labels. Through large-scale experiments across multiple LLM families and detection methods, we reveal three key findings: (1) No paradigm is universally superior; (2) HD and FV capture complementary facets of factual errors; and (3) hybrid approaches that integrate both methods consistently achieve state-of-the-art performance. Beyond benchmarking, we provide the first in-depth analysis of why FV and HD diverged, as well as empirical evidence supporting the need for their unification. The comprehensive experimental results call for a new, integrated research agenda toward unifying Hallucination Detection and Fact Verification in LLMs. We have open-sourced all the code, data, and baseline implementation at: https://github.com/oneal2000/UniFact/
Read the full abstract

All authors Weihang Su, Jianming Long, Changyue Wang, Shiyu Lin, Jingyan Xu, Ziyi Ye, Qingyao Ai, Yiqun Liu

Computation and Language

PEFT-Factory: Unified Parameter-Efficient Fine-Tuning of Autoregressive Large Language Models

Robert Belanec, Ivan Srba and Maria Bielikova

Parameter-Efficient Fine-Tuning (PEFT) methods address the increasing size of Large Language Models (LLMs). Currently, many newly introduced PEFT methods are challenging to replicate, deploy, or compare with one another. To address this, we introduce PEFT-Factory, a unified framework for efficient fine-tuning LLMs using both off-the-shelf and custom PEFT methods. While its modular design supports extensibility, it natively provides a representative set of 19 PEFT methods, 27 classification and text generation datasets addressing 12 tasks, and both standard and PEFT-specific evaluation metrics. As a result, PEFT-Factory provides a ready-to-use, controlled, and stable environment, improving replicability and benchmarking of PEFT methods. PEFT-Factory is a downstream framework that originates from the popular LLaMA-Factory, and is publicly available at https://github.com/kinit-sk/PEFT-Factory.
Read the full abstract

Computation and Language

SurveyEval: Towards Comprehensive Evaluation of LLM-Generated Academic Surveys

Jiahao Zhao, Shuaixing Zhang, Nan Xu et al. (+1 more)

LLM-based automatic survey systems are transforming how users acquire information from the web by integrating retrieval, organization, and content synthesis into end-to-end generation pipelines. While recent works focus on developing new generation pipelines, how to evaluate such complex systems remains a significant challenge. To this end, we introduce SurveyEval, a comprehensive benchmark that evaluates automatically generated surveys across three dimensions: overall quality, outline coherence, and reference accuracy. We extend the evaluation across 7 subjects and augment the LLM-as-a-Judge framework with human references to strengthen evaluation-human alignment. Evaluation results show that while general long-text or paper-writing systems tend to produce lower-quality surveys, specialized survey-generation systems are able to deliver substantially higher-quality results. We envision SurveyEval as a scalable testbed to understand and improve automatic survey systems across diverse subjects and evaluation criteria.
Read the full abstract

All authors Jiahao Zhao, Shuaixing Zhang, Nan Xu, Lei Wang

Computation and Language · Cross-listed from cs.PL

Probabilistic energy profiler for statically typed JVM-based programming languages

Joel Nyholm, Wojciech Mostowski and Christoph Reichenbach

Energy consumption is a growing concern in several fields, from mobile devices to large data centers. Developers need detailed data on the energy consumption of their software to mitigate consumption issues. Previous approaches have a broader focus, such as on specific functions or programs, rather than source code statements. They primarily focus on estimating the CPU's energy consumption using point estimates, thereby disregarding other hardware effects and limiting their use for statistical reasoning and explainability. We developed a novel methodology to address the limitations of measuring only the CPU's consumption and using point estimates, focusing on predicting the energy usage of statically typed JVM-based programming languages, such as Java and Scala. We measure the energy consumption of Bytecode patterns, the translation from the programming language's source code statement to their Java Bytecode representation. With the energy measurements, we construct a statistical model using Bayesian statistics, which allows us to predict the energy consumption through statistical distributions and analyze individual factors. The model includes three factors we obtain statically from the code: data size, data type, operation, and one factor about the hardware platform the code executes on: device. To validate our methodology, we implemented it for Java and evaluated its energy predictions on unseen programs. We observe that all four factors are influential, notably that two devices of the same model may differ in energy consumption and that the operations and data types cause consumption differences. The experiments also show that the energy prediction of programs closely follows the program's real energy consumption, validating our approach. Our work presents a methodology for constructing an energy model that future work, such as verification tools, can use for their energy estimates.
Read the full abstract

Computation and Language

Emergent Bayesian Behaviour and Optimal Cue Combination in LLMs

Julian Ma, Jun Wang and Zafeirios Fountas

Large language models (LLMs) excel at explicit reasoning, but their implicit computational strategies remain underexplored. Decades of psychophysics research show that humans intuitively process and integrate noisy signals using near-optimal Bayesian strategies in perceptual tasks. We ask whether LLMs exhibit similar behaviour and perform optimal multimodal integration without explicit training or instruction. Adopting the psychophysics paradigm, we infer computational principles of LLMs from systematic behavioural studies. We introduce a behavioural benchmark - BayesBench: four magnitude estimation tasks (length, location, distance, and duration) over text and image, inspired by classic psychophysics, and evaluate a diverse set of nine LLMs alongside human judgments for calibration. Through controlled ablations of noise, context, and instruction prompts, we measure performance, behaviour and efficiency in multimodal cue-combination. Beyond accuracy and efficiency metrics, we introduce a Bayesian Consistency Score that detects Bayes-consistent behavioural shifts even when accuracy saturates. Our results show that while capable models often adapt in Bayes-consistent ways, accuracy does not guarantee robustness. Notably, GPT-5 Mini achieves perfect text accuracy but fails to integrate visual cues efficiently. This reveals a critical dissociation between capability and strategy, suggesting accuracy-centric benchmarks may over-index on performance while missing brittle uncertainty handling. These findings reveal emergent principled handling of uncertainty and highlight the correlation between accuracy and Bayesian tendencies. We release our psychophysics benchmark and consistency metric (https://bayes-bench.github.io) as evaluation tools and to inform future multimodal architecture designs.
Read the full abstract

Computation and Language

CREST: Universal Safety Guardrails Through Cluster-Guided Cross-Lingual Transfer

Lavish Bansal and Naman Mishra

Ensuring content safety in large language models (LLMs) is essential for their deployment in real-world applications. However, existing safety guardrails are predominantly tailored for high-resource languages, leaving a significant portion of the world's population underrepresented who communicate in low-resource languages. To address this, we introduce CREST (CRoss-lingual Efficient Safety Transfer), a parameter-efficient multilingual safety classification model that supports 100 languages with only 0.5B parameters. By training on a strategically chosen subset of only 13 high-resource languages, our model utilizes cluster-based cross-lingual transfer from a few to 100 languages, enabling effective generalization to both unseen high-resource and low-resource languages. This approach addresses the challenge of limited training data in low-resource settings. We conduct comprehensive evaluations across six safety benchmarks to demonstrate that CREST outperforms existing state-of-the-art guardrails of comparable scale and achieves competitive results against models with significantly larger parameter counts (2.5B parameters and above). Our findings highlight the limitations of language-specific guardrails and underscore the importance of developing universal, language-agnostic safety systems that can scale effectively to serve global populations.
Read the full abstract

Computation and Language

An Empirical Survey of Model Merging Algorithms for Social Bias Mitigation

Daiki Shirafuji, Tatsuhiko Saito and Yasutomo Kimura

Large language models (LLMs) are known to inherit and even amplify societal biases present in their pre-training corpora, threatening fairness and social trust. To address this issue, recent work has explored “editing” LLM parameters to mitigate social bias with model merging approaches; however, there is no empirical comparison. In this work, we empirically survey seven algorithms: Linear, Karcher Mean, SLERP, NuSLERP, TIES, DELLA, and Nearswap, applying 13 open weight models in the GPT, LLaMA, and Qwen families. We perform a comprehensive evaluation using three bias datasets (BBQ, BOLD, and HONEST) and measure the impact of these techniques on LLM performance in downstream tasks of the SuperGLUE benchmark. We find a trade-off between bias reduction and downstream performance: methods achieving greater bias mitigation degrade accuracy, particularly on tasks requiring reading comprehension and commonsense and causal reasoning. Among the merging algorithms, Linear, SLERP, and Nearswap consistently reduce bias while maintaining overall performance, with SLERP at moderate interpolation weights emerging as the most balanced choice. These results highlight the potential of model merging algorithms for bias mitigation, while indicating that excessive debiasing or inappropriate merging methods may lead to the degradation of important linguistic abilities.
Read the full abstract

Computation and Language

Input Order Shapes LLM Semantic Alignment in Multi-Document Summarization

Jing Ma

Large language models (LLMs) are now used in settings such as Google's AI Overviews, where it summarizes multiple long documents. However, it remains unclear whether they weight all inputs equally. Focusing on abortion-related news, we construct 40 pro-neutral-con article triplets, permute each triplet into six input orders, and prompt Gemini 2.5 Flash to generate a neutral overview. We evaluate each summary against its source articles using ROUGE-L (lexical overlap), BERTScore (semantic similarity), and SummaC (factual consistency). One-way ANOVA reveals a significant primacy effect for BERTScore across all stances, indicating that summaries are more semantically aligned with the first-seen article. Pairwise comparisons further show that Position 1 differs significantly from Positions 2 and 3, while the latter two do not differ from each other, confirming a selective preference for the first document. The findings present risks for applications that rely on LLM-generated overviews and for agentic AI systems, where the steps involving LLMs can disproportionately influence downstream actions.
Read the full abstract

Computation and Language

Spoken Conversational Agents with Large Language Models

Chao-Han Huck Yang, Andreas Stolcke and Larry Heck

Spoken conversational agents are converging toward voice-native LLMs. This tutorial distills the path from cascaded ASR/NLU to end-to-end, retrieval-and vision-grounded systems. We frame adaptation of text LLMs to audio, cross-modal alignment, and joint speech-text training; review datasets, metrics, and robustness across accents and compare design choices (cascaded vs. E2E, post-ASR correction, streaming). We link industrial assistants to current open-domain and task-oriented agents, highlight reproducible baselines, and outline open problems in privacy, safety, and evaluation. Attendees leave with practical recipes and a clear systems-level roadmap.
Read the full abstract

Computation and Language

From Imitation to Discrimination: Toward A Generalized Curriculum Advantage Mechanism Enhancing Cross-Domain Reasoning Tasks

Changpeng Yang, Jinyang Wu, Yuchen Liu et al. (+9 more)

Reinforcement learning has emerged as a paradigm for post-training large language models, boosting their reasoning capabilities. Such approaches compute an advantage value for each sample, reflecting better or worse performance than expected, thereby yielding both positive and negative signals for training. However, the indiscriminate mixing of the two signals in existing methods, especially from the early stages, may lead to ambiguous guidance and limited gains. To address this issue, we propose **CAPO** (**C**urriculum **A**dvantage **P**olicy **O**ptimization), an adaptive curriculum mechanism based on advantage signals. The proposed mechanism bootstraps imitation learning with positive-only advantage samples to establish robust foundations, and subsequently introduces negative signals to cultivate discriminative capabilities, thereby improving generalization across complex scenarios. Compatible with diverse optimization methods including GRPO, PPO, RLOO, and Reinforce++, our method consistently achieves stable and significant improvements in mathematical reasoning tasks, and further generalizes effectively to multimodal Graphical User Interface (GUI) reasoning scenarios, establishing itself as a versatile and robust optimization framework.
Read the full abstract

All authors Changpeng Yang, Jinyang Wu, Yuchen Liu, Shuai Zhang, Yang Li, Qiliang Liang, Hongzhen Wang, Shuai Nie, Jiaming Xu, Runyu Shi, Ying Huang, Guoquan Zhang

Computation and Language

DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models

DeepSeek-AI, Aixin Liu, Aoxue Mei et al. (+261 more)

We introduce DeepSeek-V3.2, a model that harmonizes high computational efficiency with superior reasoning and agent performance. The key technical breakthroughs of DeepSeek-V3.2 are as follows: (1) DeepSeek Sparse Attention (DSA): We introduce DSA, an efficient attention mechanism that substantially reduces computational complexity while preserving model performance in long-context scenarios. (2) Scalable Reinforcement Learning Framework: By implementing a robust reinforcement learning protocol and scaling post-training compute, DeepSeek-V3.2 performs comparably to GPT-5. Notably, our high-compute variant, DeepSeek-V3.2-Speciale, surpasses GPT-5 and exhibits reasoning proficiency on par with Gemini-3.0-Pro, achieving gold-medal performance in both the 2025 International Mathematical Olympiad (IMO) and the International Olympiad in Informatics (IOI). (3) Large-Scale Agentic Task Synthesis Pipeline: To integrate reasoning into tool-use scenarios, we developed a novel synthesis pipeline that systematically generates training data at scale. This methodology facilitates scalable agentic post-training, yielding substantial improvements in generalization and instruction-following robustness within complex, interactive environments.
Read the full abstract

All authors DeepSeek-AI, Aixin Liu, Aoxue Mei, Bangcai Lin, Bing Xue, Bingxuan Wang, Bingzheng Xu, Bochao Wu, Bowei Zhang, Chaofan Lin, Chen Dong, Chengda Lu, Chenggang Zhao, Chengqi Deng, Chenhao Xu, Chong Ruan, Damai Dai, Daya Guo, Dejian Yang, Deli Chen, Erhang Li, Fangqi Zhou, Fangyun Lin, Fucong Dai, Guangbo Hao, Guanting Chen, Guowei Li, H. Zhang, Hanwei Xu, Hao Li, Haofen Liang, Haoran Wei, Haowei Zhang, Haowen Luo, Haozhe Ji, Honghui Ding, Hongxuan Tang, Huanqi Cao, Huazuo Gao, Hui Qu, Hui Zeng, Jialiang Huang, Jiashi Li, Jiaxin Xu, Jiewen Hu, Jingchang Chen, Jingting Xiang, Jingyang Yuan, Jingyuan Cheng, Jinhua Zhu, Jun Ran, Junguang Jiang, Junjie Qiu, Junlong Li, Junxiao Song, Kai Dong, Kaige Gao, Kang Guan, Kexin Huang, Kexing Zhou, Kezhao Huang, Kuai Yu, Lean Wang, Lecong Zhang, Lei Wang, Liang Zhao, Liangsheng Yin, Lihua Guo, Lingxiao Luo, Linwang Ma, Litong Wang, Liyue Zhang, M.S. Di, M.Y Xu, Mingchuan Zhang, Minghua Zhang, Minghui Tang, Mingxu Zhou, Panpan Huang, Peixin Cong, Peiyi Wang, Qiancheng Wang, Qihao Zhu, Qingyang Li, Qinyu Chen, Qiushi Du, Ruiling Xu, Ruiqi Ge, Ruisong Zhang, Ruizhe Pan, Runji Wang, Runqiu Yin, Runxin Xu, Ruomeng Shen, Ruoyu Zhang, S.H. Liu, Shanghao Lu, Shangyan Zhou, Shanhuang Chen, Shaofei Cai, Shaoyuan Chen, Shengding Hu, Shengyu Liu, Shiqiang Hu, Shirong Ma, Shiyu Wang, Shuiping Yu, Shunfeng Zhou, Shuting Pan, Songyang Zhou, Tao Ni, Tao Yun, Tian Pei, Tian Ye, Tianyuan Yue, Wangding Zeng, Wen Liu, Wenfeng Liang, Wenjie Pang, Wenjing Luo, Wenjun Gao, Wentao Zhang, Xi Gao, Xiangwen Wang, Xiao Bi, Xiaodong Liu, Xiaohan Wang, Xiaokang Chen, Xiaokang Zhang, Xiaotao Nie, Xin Cheng, Xin Liu, Xin Xie, Xingchao Liu, Xingkai Yu, Xingyou Li, Xinyu Yang, Xinyuan Li, Xu Chen, Xuecheng Su, Xuehai Pan, Xuheng Lin, Xuwei Fu, Y.Q. Wang, Yang Zhang, Yanhong Xu, Yanru Ma, Yao Li, Yao Li, Yao Zhao, Yaofeng Sun, Yaohui Wang, Yi Qian, Yi Yu, Yichao Zhang, Yifan Ding, Yifan Shi, Yiliang Xiong, Ying He, Ying Zhou, Yinmin Zhong, Yishi Piao, Yisong Wang, Yixiao Chen, Yixuan Tan, Yixuan Wei, Yiyang Ma, Yiyuan Liu, Yonglun Yang, Yongqiang Guo, Yongtong Wu, Yu Wu, Yuan Cheng, Yuan Ou, Yuanfan Xu, Yuduan Wang, Yue Gong, Yuhan Wu, Yuheng Zou, Yukun Li, Yunfan Xiong, Yuxiang Luo, Yuxiang You, Yuxuan Liu, Yuyang Zhou, Z.F. Wu, Z.Z. Ren, Zehua Zhao, Zehui Ren, Zhangli Sha, Zhe Fu, Zhean Xu, Zhenda Xie, Zhengyan Zhang, Zhewen Hao, Zhibin Gou, Zhicheng Ma, Zhigang Yan, Zhihong Shao, Zhixian Huang, Zhiyu Wu, Zhuoshu Li, Zhuping Zhang, Zian Xu, Zihao Wang, Zihui Gu, Zijia Zhu, Zilin Li, Zipeng Zhang, Ziwei Xie, Ziyi Gao, Zizheng Pan, Zongqing Yao, Bei Feng, Hui Li, J.L. Cai, Jiaqi Ni, Lei Xu, Meng Li, Ning Tian, R.J. Chen, R.L. Jin, S.S. Li, Shuang Zhou, Tianyu Sun, X.Q. Li, Xiangyue Jin, Xiaojin Shen, Xiaosha Chen, Xinnan Song, Xinyi Zhou, Y.X. Zhu, Yanping Huang, Yaohui Li, Yi Zheng, Yuchen Zhu, Yunxian Ma, Zhen Huang, Zhipeng Xu, Zhongyu Zhang, Dongjie Ji, Jian Liang, Jianzhong Guo, Jin Chen, Leyi Xia, Miaojun Wang, Mingming Li, Peng Zhang, Ruyi Chen, Shangmian Sun, Shaoqing Wu, Shengfeng Ye, T.Wang, W.L. Xiao, Wei An, Xianzu Wang, Xiaowen Sun, Xiaoxiang Wang, Ying Tang, Yukun Zha, Zekai Zhang, Zhe Ju, Zhen Zhang, Zihua Qu

Computation and Language

ADORE: Autonomous Domain-Oriented Relevance Engine for E-commerce

Zheng Fang, Donghao Xie, Ming Pang et al. (+5 more)

Relevance modeling in e-commerce search remains challenged by semantic gaps in term-matching methods (e.g., BM25) and neural models' reliance on the scarcity of domain-specific hard samples. We propose ADORE, a self-sustaining framework that synergizes three innovations: (1) A Rule-aware Relevance Discrimination module, where a Chain-of-Thought LLM generates intent-aligned training data, refined via Kahneman-Tversky Optimization (KTO) to align with user behavior; (2) An Error-type-aware Data Synthesis module that auto-generates adversarial examples to harden robustness; and (3) A Key-attribute-enhanced Knowledge Distillation module that injects domain-specific attribute hierarchies into a deployable student model. ADORE automates annotation, adversarial generation, and distillation, overcoming data scarcity while enhancing reasoning. Large-scale experiments and online A/B testing verify the effectiveness of ADORE. The framework establishes a new paradigm for resource-efficient, cognitively aligned relevance modeling in industrial applications.
Read the full abstract

All authors Zheng Fang, Donghao Xie, Ming Pang, Chunyuan Yuan, Xue Jiang, Changping Peng, Zhangang Lin, Zheng Luo

Computation and Language

From Veracity to Diffusion: Adressing Operational Challenges in Moving From Fake-News Detection to Information Disorders

Francesco Paolo Savatteri, Chahan Vidal-Gorène and Florian Cafiero

A wide part of research on misinformation has relied lies on fake-news detection, a task framed as the prediction of veracity labels attached to articles or claims. Yet social-science research has repeatedly emphasized that information manipulation goes beyond fabricated content and often relies on amplification dynamics. This theoretical turn has consequences for operationalization in applied social science research. What changes empirically when prediction targets move from veracity to diffusion? And which performance level can be attained in limited resources setups ? In this paper we compare fake-news detection and virality prediction across two datasets, EVONS and FakeNewsNet. We adopt an evaluation-first perspective and examine how benchmark behavior changes when the prediction target shifts from veracity to diffusion. Our experiments show that fake-news detection is comparatively stable once strong textual embeddings are available, whereas virality prediction is much more sensitive to operational choices such as threshold definition and early observation windows. The paper proposes practical ways to operationalize lightweight, transparent pipelines for misinformation-related prediction tasks that can rival with state-of-the-art.
Read the full abstract

Computation and Language

A Concise Review of Hallucinations in LLMs and their Mitigation

Parth Pulkundwar, Vivek Dhanawade, Rohit Yadav et al. (+3 more)

Traditional language models face a challenge from hallucinations. Their very presence casts a large, dangerous shadow over the promising realm of natural language processing. It becomes crucial to understand the various kinds of hallucinations that occur nowadays, their origins, and ways of reducing them. This document provides a concise and straightforward summary of that. It serves as a one-stop resource for a general understanding of hallucinations and how to mitigate them.
Read the full abstract

All authors Parth Pulkundwar, Vivek Dhanawade, Rohit Yadav, Minal Sonkar, Medha Asurlekar, Sarita Rathod

Computation and Language · Cross-listed from cs.AI

Guided Self-Evolving LLMs with Minimal Human Supervision

Wenhao Yu, Zhenwen Liang, Chengsong Huang et al. (+4 more)

AI self-evolution has long been envisioned as a path toward superintelligence, where models autonomously acquire, refine, and internalize knowledge from their own learning experiences. Yet in practice, unguided self-evolving systems often plateau quickly or even degrade as training progresses. These failures arise from issues such as concept drift, diversity collapse, and mis-evolution, as models reinforce their own biases and converge toward low-entropy behaviors. To enable models to self-evolve in a stable and controllable manner while minimizing reliance on human supervision, we introduce R-Few, a guided Self-Play Challenger-Solver framework that incorporates lightweight human oversight through in-context grounding and mixed training. At each iteration, the Challenger samples a small set of human-labeled examples to guide synthetic question generation, while the Solver jointly trains on human and synthetic examples under an online, difficulty-based curriculum. Across math and general reasoning benchmarks, R-Few achieves consistent and iterative improvements. For example, Qwen3-8B-Base improves by +3.0 points over R-Zero on math tasks and achieves performance on par with General-Reasoner, despite the latter being trained on 20 times more human data. Ablation studies confirm the complementary contributions of grounded challenger training and curriculum-based solver training, and further analysis shows that R-Few mitigates drift, yielding more stable and controllable co-evolutionary dynamics.
Read the full abstract

All authors Wenhao Yu, Zhenwen Liang, Chengsong Huang, Kishan Panaganti, Tianqing Fang, Haitao Mi, Dong Yu

Computation and Language · Cross-listed from cs.CV

See, Think, Learn: A Self-Taught Multimodal Reasoner

Sourabh Sharma, Sonam Gupta and Sadbhawna

Vision-Language Models (VLMs) have achieved remarkable progress in integrating visual perception with language understanding. However, effective multimodal reasoning requires both accurate perception and robust reasoning, and weakness in either limits the performance of VLMs. Prior efforts to enhance reasoning often depend on high-quality chain-of-thought (CoT) data, obtained via labor-intensive human annotations, costly proprietary models, or self-training methods that overlook perception. To address these limitations, we propose a simple yet effective self-training framework called See-Think-Learn (STL). At its core, STL introduces a structured reasoning template that encourages the model to see before thinking, first extracting visual attributes in textual form, then using them to guide reasoning. The framework jointly improves perception and reasoning by having the model generate and learn from its own structured rationales in a self-training loop. Furthermore, we augment the training data with negative rationales, i.e. explanations that justify why certain answer choices are incorrect, to enhance the model's ability to distinguish between correct and misleading responses. This fosters more discriminative and robust learning. Experiments across diverse domains show that STL consistently outperforms baselines trained directly only on answers or self-generated reasoning, while qualitative analysis confirms the high quality of its rationales. STL thus provides a cost-effective solution to enhance multimodal reasoning ability of VLMs.
Read the full abstract

Computation and Language · Cross-listed from cs.LG

When Refusals Fail: Unstable Safety Mechanisms in Long-Context LLM Agents

Tsimur Hadeliya, Mohammad Ali Jauhar, Nidhi Sakpal et al. (+1 more)

Solving complex or long-horizon problems often requires large language models (LLMs) to use external tools and operate over a significantly longer context window. New LLMs enable longer context windows and support tool calling capabilities. Prior works have focused mainly on evaluation of LLMs on long-context prompts, leaving agentic setup relatively unexplored, both from capability and safety perspectives. Our work addresses this gap. We find that LLM agents could be sensitive to length, type, and placement of the context, exhibiting unexpected and inconsistent shifts in task performance and in refusals to execute harmful requests. Models with 1M-2M token context windows show severe degradation already at 100K tokens, with performance drops exceeding 50% for both benign and harmful tasks. Refusal rates shift unpredictably: GPT-4.1-nano increases from $\sim$5% to $\sim$40% while Grok 4 Fast decreases from $\sim$80% to $\sim$10% at 200K tokens. Our work shows potential safety issues with agents operating on longer context and opens additional questions on the current metrics and paradigm for evaluating LLM agent safety on long multi-step tasks. In particular, our results on LLM agents reveal a notable divergence in both capability and safety performance compared to prior evaluations of LLMs on similar criteria.
Read the full abstract

All authors Tsimur Hadeliya, Mohammad Ali Jauhar, Nidhi Sakpal, Diogo Cruz

Computation and Language · Cross-listed from cs.CV

WorldMM: Dynamic Multimodal Memory Agent for Long Video Reasoning

Woongyeong Yeo, Kangsan Kim, Jaehong Yoon et al. (+1 more)

Recent advances in video large language models have demonstrated strong capabilities in understanding short clips. However, scaling them to hours- or days-long videos remains highly challenging due to limited context capacity and the loss of critical visual details during abstraction. Existing memory-augmented methods mitigate this by leveraging textual summaries of video segments, yet they heavily rely on text and fail to utilize visual evidence when reasoning over complex scenes. Moreover, retrieving from fixed temporal scales further limits their flexibility in capturing events that span variable durations. To address this, we introduce WorldMM, a novel multimodal memory agent that constructs and retrieves from multiple complementary memories, encompassing both textual and visual representations. WorldMM comprises three types of memory: episodic memory indexes factual events across multiple temporal scales, semantic memory continuously updates high-level conceptual knowledge, and visual memory preserves detailed information about scenes. During inference, an adaptive retrieval agent iteratively selects the most relevant memory source and leverages multiple temporal granularities based on the query, continuing until it determines that sufficient information has been gathered. WorldMM significantly outperforms existing baselines across five long video question-answering benchmarks, achieving an average 8.4% performance gain over previous state-of-the-art methods, showing its effectiveness on long video reasoning.
Read the full abstract

All authors Woongyeong Yeo, Kangsan Kim, Jaehong Yoon, Sung Ju Hwang

Computation and Language · Cross-listed from q-bio.NC

The brain-AI convergence: Predictive and generative world models for general-purpose computation

Shogo Ohmae and Keiko Ohmae

Recent advances in general-purpose AI systems with attention-based transformers offer a potential window into how the neocortex and cerebellum, despite their relatively uniform circuit architectures, give rise to diverse functions and, ultimately, to human intelligence. This Perspective provides a cross-domain comparison between the brain and AI that goes beyond the traditional focus on visual processing, adopting the emerging perspecive of world-model-based computation. Here, we identify shared computational mechanisms in the attention-based neocortex and the non-attentional cerebellum: both predict future world events from past inputs and construct internal world models through prediction-error learning. These predictive world models are repurposed for seemingly distinct functions – understanding in sensory processing and generation in motor processing – enabling the brain to achieve multi-domain capabilities and human-like adaptive intelligence. Notably, attention-based AI has independently converged on a similar learning paradigm and world-model-based computation. We conclude that these shared mechanisms in both biological and artificial systems constitute a core computational foundation for realizing diverse functions including high-level intelligence, despite their relatively uniform circuit structures. Our theoretical insights bridge neuroscience and AI, advancing our understanding of the computational essence of intelligence.
Read the full abstract

Computation and Language

TaleFrame: An Interactive Story Generation System with Fine-Grained Control and Large Language Models

Yunchao Wang, Guodao Sun, Zihang Fu et al. (+4 more)

With the advancement of natural language generation (NLG) technologies, creative story generation systems have gained increasing attention. However, current systems often fail to accurately translate user intent into satisfactory story outputs due to a lack of fine-grained control and unclear input specifications, limiting their applicability. To address this, we propose TaleFrame, a system that combines large language models (LLMs) with human-computer interaction (HCI) to generate stories through structured information, enabling precise control over the generation process. The innovation of TaleFrame lies in decomposing the story structure into four basic units: entities, events, relationships, and story outline. We leverage the Tinystories dataset, parsing and constructing a preference dataset consisting of 9,851 JSON-formatted entries, which is then used to fine-tune a local Llama model. By employing this JSON2Story approach, structured data is transformed into coherent stories. TaleFrame also offers an intuitive interface that supports users in creating and editing entities and events and generates stories through the structured framework. Users can control these units through simple interactions (e.g., drag-and-drop, attach, and connect), thus influencing the details and progression of the story. The generated stories can be evaluated across seven dimensions (e.g., creativity, structural integrity), with the system providing suggestions for refinement based on these evaluations. Users can iteratively adjust the story until a satisfactory result is achieved. Finally, we conduct quantitative evaluation and user studies that demonstrate the usefulness of TaleFrame. Dataset available at https://huggingface.co/datasets/guodaosun/tale-frame.
Read the full abstract

All authors Yunchao Wang, Guodao Sun, Zihang Fu, Zhehao Liu, Kaixing Du, Haidong Gao, Ronghua Liang

Computation and Language · Cross-listed from cs.SE

Process-Centric Analysis of Agentic Software Systems

Shuyang Liu, Yang Chen, Rahul Krishna et al. (+3 more)

Agentic systems are modern software systems: they consist of orchestrated modules, expose interfaces, and are deployed in software pipelines. Unlike conventional programs, their execution, i.e., trajectories, is inherently stochastic and adaptive to the problems they solve. Evaluation of such systems is often outcome-centric. This narrow focus overlooks detailed insights, failing to explain how agents reason, plan, act, or change their strategies. Inspired by the structured representation of conventional software systems as graphs, we introduce Graphectory to systematically encode the temporal and semantic relations in such systems. Using Graphectory, we automatically analyze 4000 trajectories of two dominant agentic programming workflows, SWE-agent and OpenHands, with four backbone Large Language Models (LLMs), attempting to resolve SWE-bench issues. Our automated analyses (completed within four minutes) reveal that: (1) agents using richer prompts or stronger LLMs exhibit more complex Graphectory, reflecting deeper exploration, broader context gathering, and more thorough validation; (2) agents' strategies vary with problem difficulty and the underlying LLM - for resolved issues, strategies often follow coherent localization-patching-validation steps, while unresolved ones exhibit chaotic or backtracking behaviors; and (3) even successful agentic systems often display inefficient processes. We also implement a novel technique for real-time construction and analysis of Graphectory and Langutory during agent execution to flag trajectory issues. Upon detecting such issues, the technique notifies the agent with a diagnostic message and, when applicable, rolls back the trajectory. Experiments show that online monitoring and interventions improve resolution rates by 6.9%-23.5% across models for problematic instances, while significantly shortening trajectories with near-zero overhead.
Read the full abstract

All authors Shuyang Liu, Yang Chen, Rahul Krishna, Saurabh Sinha, Jatin Ganhotra, Reyhan Jabbarvand

Computation and Language

Memory-Augmented Knowledge Fusion with Safety-Aware Decoding for Domain-Adaptive Question Answering

Lei Fu, Xiang Chen, Kaige Gao Xinyue Huang et al. (+1 more)

Domain-specific question answering (QA) systems for services face unique challenges in integrating heterogeneous knowledge sources while ensuring both accuracy and safety. Existing large language models often struggle with factual consistency and context alignment in sensitive domains such as healthcare policies and government welfare. In this work, we introduce Knowledge-Aware Reasoning and Memory-Augmented Adaptation (KARMA), a novel framework designed to enhance QA performance in care scenarios. KARMA incorporates a dual-encoder architecture to fuse structured and unstructured knowledge sources, a gated memory unit to dynamically regulate external knowledge integration, and a safety-aware controllable decoder that mitigates unsafe outputs using safety classification and guided generation techniques. Extensive experiments on a proprietary QA dataset demonstrate that KARMA outperforms strong baselines in both answer quality and safety. This study offers a comprehensive solution for building trustworthy and adaptive QA systems in service contexts.
Read the full abstract

All authors Lei Fu, Xiang Chen, Kaige Gao Xinyue Huang, Kejian Tong

Computation and Language · Cross-listed from cs.CR

LeechHijack: Covert Computational Resource Exploitation in Intelligent Agent Systems

Yuanhe Zhang, Weiliu Wang, Zhenhong Zhou et al. (+5 more)

Large Language Model (LLM)-based agents have demonstrated remarkable capabilities in reasoning, planning, and tool usage. The recently proposed Model Context Protocol (MCP) has emerged as a unifying framework for integrating external tools into agent systems, enabling a thriving open ecosystem of community-built functionalities. However, the openness and composability that make MCP appealing also introduce a critical yet overlooked security assumption – implicit trust in third-party tool providers. In this work, we identify and formalize a new class of attacks that exploit this trust boundary without violating explicit permissions. We term this new attack vector implicit toxicity, where malicious behaviors occur entirely within the allowed privilege scope. We propose LeechHijack, a Latent Embedded Exploit for Computation Hijacking, in which an adversarial MCP tool covertly expropriates the agent's computational resources for unauthorized workloads. LeechHijack operates through a two-stage mechanism: an implantation stage that embeds a benign-looking backdoor in a tool, and an exploitation stage where the backdoor activates upon predefined triggers to establish a command-and-control channel. Through this channel, the attacker injects additional tasks that the agent executes as if they were part of its normal workflow, effectively parasitizing the user's compute budget. We implement LeechHijack across four major LLM families. Experiments show that LeechHijack achieves an average success rate of 77.25%, with a resource overhead of 18.62% compared to the baseline. This study highlights the urgent need for computational provenance and resource attestation mechanisms to safeguard the emerging MCP ecosystem.
Read the full abstract

All authors Yuanhe Zhang, Weiliu Wang, Zhenhong Zhou, Kun Wang, Jie Zhang, Li Sun, Yang Liu, Sen Su

Computation and Language · Cross-listed from cs.AI

OmniGuard: Unified Omni-Modal Guardrails with Deliberate Reasoning

Boyu Zhu, Xiaofei Wen, Wenjie Jacky Mo et al. (+4 more)

Omni-modal Large Language Models (OLLMs) that process text, images, videos, and audio introduce new challenges for safety and value guardrails in human-AI interaction. Prior guardrail research largely targets unimodal settings and typically frames safeguarding as binary classification, which limits robustness across diverse modalities and tasks. To address this gap, we propose OmniGuard, the first family of omni-modal guardrails that performs safeguarding across all modalities with deliberate reasoning ability. To support the training of OMNIGUARD, we curate a large, comprehensive omni-modal safety dataset comprising over 210K diverse samples, with inputs that cover all modalities through both unimodal and cross-modal samples. Each sample is annotated with structured safety labels and carefully curated safety critiques from expert models through targeted distillation. Extensive experiments on 15 benchmarks show that OmniGuard achieves strong effectiveness and generalization across a wide range of multimodal safety scenarios. Importantly, OmniGuard provides a unified framework that enforces policies and mitigates risks in omni-modalities, paving the way toward building more robust and capable omnimodal safeguarding systems.
Read the full abstract

All authors Boyu Zhu, Xiaofei Wen, Wenjie Jacky Mo, Tinghui Zhu, Yanan Xie, Peng Qi, Muhao Chen

Computation and Language

When Does Verification Pay Off? A Closer Look at LLMs as Solution Verifiers

Jack Lu, Ryan Teehan, Jinran Jin et al. (+1 more)

Large language models (LLMs) can act as both problem solvers and solution verifiers, where the latter select high-quality answers from a pool of solver-generated candidates. This raises the question of under what conditions verification pays off in solver-verifier systems. Prior work has conducted only limited studies of the factors influencing verification performance, focusing primarily on self-verification and examining neither the relationship between solver and verifier model families nor the effects of reasoning post-training. To rectify this, we present a systematic study across 37 models spanning multiple families, sizes, and base vs. post-trained variants, evaluated on 9 benchmarks covering logical reasoning, structured puzzles, symbolic computation, mathematics, commonsense, factual recall, and domain knowledge. In order to support our analysis, we introduce and empirically validate verifier gain, a metric that predicts the performance improvements from test-time verifier-based rejection sampling. Our experiments find that 1) verification across model families is more effective than either self-verification or verification within the same family, and more generally that the benefits of verification decrease as the solver and verifier become more similar, 2) reasoning post-training weakens self-improvement abilities but strengthens cross-family improvement, and 3) some tasks are inherently more amenable to improvement through verification, particularly mathematical and logical tasks.
Read the full abstract

All authors Jack Lu, Ryan Teehan, Jinran Jin, Mengye Ren

Computation and Language

HealthContradict: Evaluating Biomedical Knowledge Conflicts in Language Models

Boya Zhang, Alban Bornet, Rui Yang et al. (+2 more)

How do language models use contextual information to answer health questions? How are their responses impacted by conflicting contexts? We assess the ability of language models to reason over long, conflicting biomedical contexts using HealthContradict, an expert-verified dataset comprising 920 unique instances, each consisting of a health-related question, a factual answer supported by scientific evidence, and two documents presenting contradictory stances. We consider several prompt settings, including correct, incorrect or contradictory context, and measure their impact on model outputs. Compared to existing medical question-answering evaluation benchmarks, HealthContradict provides greater distinctions of language models' contextual reasoning capabilities. Our experiments show that the strength of fine-tuned biomedical language models lies not only in their parametric knowledge from pretraining, but also in their ability to exploit correct context while resisting incorrect context.
Read the full abstract

All authors Boya Zhang, Alban Bornet, Rui Yang, Nan Liu, Douglas Teodoro

Monday 1 December

Computation and Language · Cross-listed from cs.HC

Serendipity with Generative AI: Repurposing knowledge components during polycrisis with a Viable Systems Model approach

Gordon Fletcher and Saomai Vu Khan

Organisations face polycrisis uncertainty yet overlook embedded knowledge. We show how generative AI can operate as a serendipity engine and knowledge transducer to discover, classify and mobilise reusable components (models, frameworks, patterns) from existing documents. Using 206 papers, our pipeline extracted 711 components (approx 3.4 per paper) and organised them into a repository aligned to Beer's Viable System Model (VSM). We contribute i) conceptually, a theory of planned serendipity in which GenAI lowers transduction costs between VSM subsystems, ii) empirically, a component repository and temporal/subject patterns, iii) managerially, a vignette and process blueprint for organisational adoption and iv) socially, pathways linking repurposing to environmental and social benefits. We propose testable links between repository creation, discovery-to-deployment time, and reuse rates, and discuss implications for shifting innovation portfolios from breakthrough bias toward systematic repurposing.
Read the full abstract

Computation and Language · Cross-listed from cs.LG

Call2Instruct: Automated Pipeline for Generating Q&A Datasets from Call Center Recordings for LLM Fine-Tuning

Alex Echeverria, Sávio Salvarino Teles de Oliveira and Fernando Marques Federson

The adaptation of Large-Scale Language Models (LLMs) to specific domains depends on high-quality fine-tuning datasets, particularly in instructional format (e.g., Question-Answer - Q&A). However, generating these datasets, particularly from unstructured sources such as call center audio recordings, poses a significant challenge due to the noisy and disorganized nature of the data. This paper presents a solution to this challenge by offering an end-to-end automated pipeline for generating Q&A instructional datasets from such recordings. The methodology developed comprises sequential steps of audio processing (including diarization, noise removal and automatic transcription), textual processing (cleaning, normalization, and anonymization), semantic extraction of customer demands and attendant responses using vector embeddings, and matching via semantic search to form the final Q&A pairs. As a result, the complete pipeline was successfully implemented, generating a dataset specifically formatted for Instruct Fine Tuning. The practical value and feasibility of the generated dataset were substantiated and functionally demonstrated through the successful fine-tuning of an LLM model (based on Llama 2 7B). The conclusion of the paper states that the proposed approach is viable for converting unstructured conversational data from call centers into valuable resources for training LLMs. This development has the potential to open up avenues for creating more effective AI systems for Q&A tasks in the customer service domain. The developed codes have been made publicly available to promote reproducibility and future research.
Read the full abstract

Computation and Language

Enhancing Foundation Models in Transaction Understanding with LLM-based Sentence Embeddings

Xiran Fan, Zhimeng Jiang, Chin-Chia Michael Yeh et al. (+4 more)

The ubiquity of payment networks generates vast transactional data encoding rich consumer and merchant behavioral patterns. Recent foundation models for transaction analysis process tabular data sequentially but rely on index-based representations for categorical merchant fields, causing substantial semantic information loss by converting rich textual data into discrete tokens. While Large Language Models (LLMs) can address this limitation through superior semantic understanding, their computational overhead challenges real-time financial deployment. We introduce a hybrid framework that uses LLM-generated embeddings as semantic initializations for lightweight transaction models, balancing interpretability with operational efficiency. Our approach employs multi-source data fusion to enrich merchant categorical fields and a one-word constraint principle for consistent embedding generation across LLM architectures. We systematically address data quality through noise filtering and context-aware enrichment. Experiments on large-scale transaction datasets demonstrate significant performance improvements across multiple transaction understanding tasks.
Read the full abstract

All authors Xiran Fan, Zhimeng Jiang, Chin-Chia Michael Yeh, Yuzhong Chen, Yingtong Dou, Menghai Pan, Yan Zheng

Computation and Language

KVReviver: Reversible KV Cache Compression with Sketch-Based Token Reconstruction

Aomufei Yuan, Zhiming Wang, Ruijie Miao et al. (+8 more)

As the context length of current large language models (LLMs) rapidly increases, the memory demand for the Key-Value (KV) cache is becoming a bottleneck for LLM deployment and batch processing. Traditional KV cache compression methods typically involve permanently evicting or irreversibly merging "less important" tokens with low attention scores. This approach results in the unrecoverable loss of token information, which we call Contextual Amnesia, significantly degrading the model's information retrieval capability. To address this issue, we propose KVReviver, a reversible KV cache compression method based on the sketch algorithm. This method allows reconstructing compressed tokens from an additional data structure, thus enabling full-scale computation within limited memory. Experiments showed that in 2k-length contexts, it requires only 10% of KV Cache budget while maintaining identical end-to-end inference accuracy. For 32k-length contexts, it achieves equivalent or comparable accuracy ~2% accuracy loss) using merely 25% of KV Cache budget.
Read the full abstract

All authors Aomufei Yuan, Zhiming Wang, Ruijie Miao, Dayu Wang, Yuxuan Tian, Zihan Wang, Yebo Peng, Yuhan Wu, Bairen Yi, Xin Liu, Tong Yang