论文检索

Zheng Luo, T Pranav Kutralingam, Ogochukwu N. Okoani, Wanpeng Xu, Hua Wei, Xiyang Hu

Large Language Models (LLMs) are increasingly deployed as agents that invoke external tools through structured function calls. While recent work reports strong tool-calling performance under standard English-centric evaluations, the robustness of tool calling under multilingual user interactions remains underexplored. In this work, we introduce MLCL, a diagnostic benchmark, and conduct a systematic evaluation of multilingual tool calling across Chinese, Hindi, and the low-resource language Igbo. Through fine-grained error analysis, we show that many failures occur despite correct intent understanding and tool selection. We identify parameter value language mismatch as a dominant failure mode, where models generate semantically appropriate parameter values in the user’s language, violating language-invariant execution conventions. We further evaluate several inference-time system strategies and find that while these strategies substantially reduce language-induced execution errors, none of them can fully recover English-level performance.

Ruifeng Yuan, Wanxing Chang, Weiwei Cao, Bowen Shi, Zhongyu Wei, Ling Zhang, Jianpeng Zhang

The evaluation of generated reports remains a critical challenge in Computed Tomography (CT) report generation, due to the large volume of text, the diversity and complexity of findings, and the presence of fine-grained, disease-oriented attributes. Conventional evaluation metrics offer only coarse measures of lexical overlap or entity matching and fail to reflect the granular diagnostic accuracy required for clinical use. To address this gap, we propose CT-FineBench, a benchmark built from CT-RATE and Merlin to evaluate the fine-grained factual consistency of CT reports, constructed from CT-RATE and Merlin. Our benchmark is constructed through a meticulous, Question-Answering (QA) based process: first, we identify and structure key, finding-specific clinical attributes (e.g., location, size, margin). Second, we systematically transform these attributes into a QA dataset, where questions probe for specific clinical details grounded in gold-standard reports. The evaluation protocol for CT-FineBench involves using this QA dataset to query a machine-generated report and scoring the correctness of the answers. This allows for a comprehensive, interpretable, and clinically-relevant assessment, moving beyond superficial lexical overlap to pinpoint specific clinical errors. Experiments show that CT-FineBench correlates better with expert clinical assessment and is substantially more sensitive to fine-grained factual errors than prior metrics.

Ziyi Wang, Yuxuan Lu, Yimeng Zhang, Pei Chen, Ziwei Dong, Jing Huang, Jiri Gesi, Xianfeng Tang, Chen Luo, Qun Liu 等

Tool-calling agents are increasingly deployed in real-world customer-facing workflows. Yet most studies on tool-calling agents focus on idealized settings with general, fixed, and well-specified tasks.In real-world applications, user requests are often (1) ambiguous, (2) changing over time, or (3) infeasible due to policy constraints, and training and evaluation data that cover these diverse, complex interaction patterns remain under-represented.To bridge the gap, we present Trajectory2Task a verifiable data generation pipeline for studying tool use at scale under three realistic user scenarios: ambiguous intent, changing intent, and infeasible intents.The pipeline first conducts multi-turn exploration to produce valid tool-call trajectories. It then converts these trajectories into user-facing tasks with controlled intent adaptations. This process yields verifiable task that support closed-loop evaluation and training. We benchmark several state-of-the-art LLMs on the generated complex user scenario tasks and observe frequent failures.Finally, using successful trajectories obtained from task rollouts, we fine-tune lightweight LLMs and find consistent improvements across all three conditions, along with better generalization to unseen tool-use domains, indicating stronger tool-calling ability.

Qingyu Lu, Liang Ding, Kanjian Zhang, Jinxia Zhang, Dacheng Tao

The pursuit of real-time agentic interaction has driven interest in Diffusion-based Large Language Models (dLLMs) as alternatives to auto-regressive backbones, promising to break the sequential latency bottleneck. **However, does such efficiency gains translate into effective agentic behavior?** In this work, we present a comprehensive evaluation of dLLMs (e.g., LLaDA, Dream) across two distinct agentic paradigms: Embodied Agents (requiring long-horizon planning) and Tool-Calling Agents (requiring precise formatting).Contrary to the efficiency hype, our results on Agentboard and BFCL reveal a "**bitter lesson**": current dLLMs fail to serve as reliable agentic backbones, frequently leading to systematically failure. **(1) In Embodied settings**, dLLMs suffer repeated attempts, failing to branch under temporal feedback. **(2) In Tool-Calling settings**, dLLMs fail to maintain symbolic precision (e.g. strict JSON schemas) under diffusion noise. To assess the potential of dLLMs in agentic workflows, we introduce **DiffuAgent**, a multi-agent evaluation framework that integrates dLLMs as plug-and-play cognitive cores. Our analysis shows that dLLMs are effective in non-causal roles (e.g., memory summarization and tool selection) but require the incorporation of causal, precise, and logically grounded reasoning mechanisms into the denoising process to be viable for agentic tasks.

Weicong Liu, Zixuan Yang, Yibo Zhao, Xiang Li

Reviewer assignment is increasingly critical yet challenging in the LLM era, where rapid topic shifts render many pre-2023 benchmarks outdated and where proxy signals poorly reflect true reviewer familiarity. We address this evaluation bottleneck by introducing LR-bench, a high-fidelity, up-to-date benchmark curated from 2024–2025 AI/NLP manuscripts with five-level self-assessed familiarity ratings collected via a large-scale email survey, yielding 1,055 expert-annotated paper–reviewer–score annotations. We further propose a reviewer-centric ranking framework that distills each reviewer’s recent publications into compact keyword-based profiles and fine-tunes an embedding model with weak preference supervision constructed from heuristic retrieval signals, enabling to match each manuscript against a reviewer profile directly. Across the LR-bench and the CMU gold-standard dataset, our approach consistently achieves state-of-the-art performance, outperforming strong embedding baselines by a clear margin. We release LR-bench at https://huggingface.co/datasets/Gnociew/LR-bench, and an github repository at https://github.com/Gnociew/RATE-Reviewer-Assignment.

Yuxuan Lu, Jing Huang, Yan Han, Bingsheng Yao, Sisong Bei, Yaochen Xie, Yisi Sang, Qi He, Dakuo Wang

Recent research shows that LLM Agents can generate “believable” human behaviors via prompt-only methods, and such agents have been increasingly adopted in downstream applications. However, existing evaluation of these agents only focuses on qualitative believability (whether human raters think they are accurate), leaving open questions of whether LLM agents can accurately generate step-by-step actions mimicking a particular human’s behavior in a multi-turn interaction task. In this work, we take shopping as a case study and present the first large-scale quantitative evaluation of state-of-the-art LLMs’ ability to accurately simulate human behavior. Using real-world data from 31,865 online shopping sessions containing 230,965 user actions, our evaluation reveals that prompt-based LLMs (DeepSeek-R1, Llama, Claude) achieve only 11.86% accuracy in generating human actions, highlighting a substantial gap in actual behavioral accuracy. Through experiments, we also showcase that strategies as simple as fine-tuning LLMs on real human click-through data augmented with synthesized reasoning traces can greatly enhance models’ performance. The fine-tuned Qwen2.5-7B achieves 17.26% action generation accuracy and 33.86% F1 score on final purchase prediction, representing substantial improvements of 5.4% and 13.85% over prompt-only baselines. This work establishes the first rigorous benchmark and dataset for human behavior simulation and provides actionable insights for developing more accurate LLM agents for future downstream applications.

Ziyi Wang, Yuxuan Lu, Wenbo Li, Amirali Amini, Bo Sun, Yakov Bart, Weimin Lyu, Jiri Gesi, Tian Wang, Jing Huang 等

Can Large Language models (LLMs) accurately simulate the next web action of a specific user? While LLMs have shown promising capabilities in generating believable human behaviors, evaluating their ability to mimic real user behaviors remains an open challenge, largely due to the lack of high-quality, publicly available datasets that capture both the observable actions and the internal reasoning of an actual human user. To address this gap, we introduce OPeRA, a novel dataset of Observation, Persona, Rationale, and Action collected from real human participants during online shopping sessions. **OPeRA is the first public dataset that comprehensively captures: user personas, browser observations, fine-grained web actions, and self-reported just-in-time rationales**. We developed both an online questionnaire and a custom browser plugin to gather this dataset with high fidelity. Using OPeRA, we establish **the first benchmark to evaluate how well current LLMs can predict a specific user’s next action** and rationale with a given persona and <observation, action, rationale> history. This dataset lays the groundwork for future research into LLM agents that aim to act as personalized digital twins for human.

Byeongjin Kim, Gyuwan Kim, Seo Yeon Park

Large language models struggle with reasoning over long contexts where relevant information is sparsely distributed. Although plan-and-execute frameworks mitigate this by decomposing tasks into planning and execution, their effectiveness is often limited by unreliable plan generation due to dependence on surface-level cues. Consequently, plans may be based on incorrect assumptions, and once a plan is formed, identifying errors and revising it reliably becomes difficult, limiting the effectiveness of reactive refinement. To address this limitation, we propose PPA-Plan, a proactive planning strategy for long-context reasoning that focuses on preventing such failures before plan generation. PPA-Plan identifies potential logical pitfalls and false assumptions, formulates them as negative constraints, and conditions plan generation on explicitly avoiding these constraints. Experiments on long-context QA benchmarks show that executing plans generated by PPA-Plan consistently outperforms existing plan-and-execute methods and direct prompting.

Lishui Fan, Yu Zhang, Mouxiang Chen, Zhongxin Liu

In practice, rigorous reasoning is often a key driver of correct code, while Reinforcement Learning (RL) for code generation often neglects optimizing reasoning quality. Bringing process-level supervision into RL is appealing, but it faces two challenges. First, training reliable reward models to assess reasoning quality is bottlenecked by the scarcity of fine-grained preference data. Second, naively incorporating such neural rewards may suffer from reward hacking. This work proposes ReCode(Reasoning-Reinforced Code Generation), a novel RL training framework comprising: (1) Contrastive Reasoning-Process Reward Learning (CRPL), which trains a reward model with synthesized optimized and degraded reasoning variants to assess the quality of reasoning process; and (2) Consistency-Gated GRPO (CG-GRPO), which integrates the reasoning-process reward model into RL by gating neural reasoning-process rewards with strict execution outcomes, using execution correctness as a hard gate to mitigate reward hacking. Additionally, to assess the reward model’s discriminative capability in assessing reasoning-process quality, we introduce LiveCodeBench-RewardBench (LCB-RB), a new benchmark comprising preference pairs of superior and inferior reasoning processes tailored for code generation. Experimental results across HumanEval(+), MBPP(+), LiveCodeBench, and BigCodeBench show that a 7B model trained with ReCode outperforms the base version by 16.1% and reaches performance comparable to GPT-4-Turbo. We further demonstrate the generalizability of ReCode by extending it to the math domain.

Connor Baumler, Calvin Bao, Huy Nghiem, Xinchen Yang, Marine Carpuat, Hal Daumé Iii

Despite the growing use of large language models (LLMs) for writing tasks, users may hesitate to rely on LLMs when personal style is important. Post-editing LLM-generated drafts or translations is a common collaborative writing strategy, but it remains unclear whether users can effectively reshape LLM-generated text to reflect their personal style. We conduct a pre-registered online study (n=81) in which participants post-edit LLM-generated drafts for writing tasks where personal style matters to them. Using embedding-based style similarity metrics, we find that post-editing increases stylistic similarity to participants’ unassisted writing and reduces similarity to fully LLM-generated output. However, post-edited text still remains stylistically closer in style to LLM text than to participants’ unassisted control text, and it exhibits reduced stylistic diversity compared to unassisted human text. We find a gap between perceived stylistic authenticity and model-measured stylistic similarity, with post-edited text often perceived as representative of participants’ personal style despite remaining detectable LLM stylistic traces.

Yuchun Fan, Bei Li, Peiguang Li, Yilin Wang, Yongyu Mu, Jian Yang, Xin Chen, Rongxiang Weng, Jingang Wang, Xunliang Cai 等

Reinforcement learning has proven effective for enhancing multi-step reasoning in Large Language Models (LLMs), yet its benefits have not fully translated to multilingual contexts. Existing methods struggle with a fundamental trade-off: prioritizing input-language consistency severely hampers reasoning quality, while prioritizing reasoning often leads to unintended language drift toward English. We address this challenge with LANG, a novel framework that leverages language-conditioned hints to guide exploration in non-English reasoning tasks. Our method incorporates two key mechanisms to prevent dependency on these hints: a progressive decay schedule that gradually withdraws scaffolding, and a language-adaptive switch that tailors learning horizons to specific language difficulties. Empirical results on challenging multilingual mathematical benchmarks reveal that LANG substantially enhances reasoning performance without compromising language consistency. Moreover, we show that our framework generalizes beyond mathematics, fostering more consistent language alignment across model layers.

Derrick Goh Xin Deik, Quanyu Long, Zhengyuan Liu, Nancy F. Chen, Wenya Wang

Multi-constraint planning involves identifying, evaluating, and refining candidate plans while satisfying multiple, potentially conflicting constraints. Existing large language model (LLM) approaches face fundamental limitations in this domain. Pure reasoning paradigms, which rely on long natural language chains, are prone to inconsistency, error accumulation, and prohibitive cost as constraints compound. Conversely, LLMs combined with coding- or solver-based strategies lack flexibility: they often generate problem-specific code from scratch or depend on fixed solvers, failing to capture generalizable logic across diverse problems.To alleviate these issues, we introduce the Scalable Code Planning Engine (SCOPE), a systematic framework that disentangles query-specific problem reasoning from generic code execution. SCOPE first transforms input queries into optimized structured representations, capturing the interdependent constraints, and then autonomously generates reusable solver functions (Combination, Filter, and Deliver) that provide consistent and reliable execution across diverse problems. SCOPE achieves state-of-the-art performance while lowering cost and latency. For example, with GPT-4o, it reaches 93.1% success on TravelPlanner, a 61.6% gain over the best baseline (CoT) while cutting inference cost by 1.4 times and time by approximately 4.67 times. Code is available at https://github.com/DerrickGXD/SCOPE.

Zelin Li, Jia Leng, Dawei Song, Yangen Hu

Direct Alignment Algorithms (DAAs) such as DPO simplify RLHF by optimizing policies directly from preference pairs. However, the Bradley–Terry probability-gap objective can induce likelihood displacement and, under weak KL constraints, may even reduce the probability of preferred responses, while implicit rewards can be limited in generalizaiton. We propose Reward Alignment Optimization (RAO), a point-wise direct alignment method that uses an explicit reward model to specify exact target generation probabilities and align the policy offline towards them. Our key insight is a theoretical principle we call "prefix consistency", which links the normalization terms of prompts that share a prefix. Leveraging this property, RAO decouples target reward differentials from bias terms, prevents decreasing preferred-response probabilities, and better exploits reward information both within and across prompts. Extensive experiments on multiple base LLMs show that RAO consistently outperforms existing DAAs while enabling controllable target probability distributions.

Ziqing Wang, Kexin Zhang, Zihan Zhao, Yibo Wen, Abhishek Pandey, Han Liu, Kaize Ding

Large language models (LLMs) are introducing a paradigm shift in molecular discovery by enabling text-guided interaction with chemical spaces through natural language and symbolic notations, with emerging extensions to incorporate multi-modal inputs. To advance this emerging field, this survey provides an up-to-date and forward-looking review of the emerging use of LLMs for two central tasks: molecule generation and molecule optimization. We organize our survey around four fundamental challenges that have emerged as critical evaluation dimensions in recent studies: ensuring validity, enhancing synthesizability, achieving precise property control, and maximizing diversity. Based on this, we systematically analyze how current LLM learning paradigms are applied to tackle each challenge, revealing the distinct capabilities and inherent limitations of each approach. In addition, we include the commonly used datasets and evaluation protocols aligned with these challenges. We conclude by discussing future directions, positioning this survey as a resource for researchers working at the intersection of LLMs and molecular science. A continuously updated reading list is available at https://github.com/REAL-Lab-NU/Awesome-LLM-Centric-Molecular-Discovery.

Lingfeng Zhong, Qiongkai Xu, Usman Naseem

Existing works on defending against LLM backdoor attacks rely on either auxiliary models or safety-related datasets for defending against backdoor attacks on large language models, which are not always available. To address these challenges, we propose our we propose our Contrastive-Selective Activation Decomposition and Steering (CS-ADS), which contrasts relatively more benign and poisoned settings to decompose the feature vectors for steering without relying on additional auxiliary models or datasets. With such disentangled vectors for remediation, our method can achieve feasible defense qualities even better than dataset-based contrastive steering strategies. This novel decomposition-based solution is motivated by the key insight that feature representations of prompt pairs can encode the same benign semantics in different proportions, even when both prompt pairs are similarly backdoored. Such discrepancies allow our method to identify effective remediation directions for steering the generation process, thereby preventing undesired outputs. We evaluate CS-ADS against multiple state-of-the-art backdoor attacks, and experimental results show that CS-ADS provides effective defense across settings.

Ziqing Wang, Yibo Wen, Abhishek Pandey, Han Liu, Kaize Ding

In drug discovery, molecular optimization aims to iteratively refine a lead compound to improve molecular properties while preserving structural similarity to the original molecule. However, each oracle evaluation is expensive, making sample efficiency a key challenge for existing methods under a limited oracle budget. Trial-and-error approaches require many oracle calls, while methods that leverage external knowledge tend to reuse familiar templates and struggle on challenging objectives. A key missing piece is long-term memory that can ground decisions and provide reusable insights for future optimizations. To address this, we present MolMem (Molecular optimization with Memory), a multi-turn agentic reinforcement learning (RL) framework with a dual-memory system. Specifically, MolMem uses Static Exemplar Memory to retrieve relevant exemplars for cold-start grounding, and Evolving Skill Memory to distill successful trajectories into reusable strategies. Built on this memory-augmented formulation, we train the policy with dense step-wise rewards, turning costly rollouts into long-term knowledge that improves future optimization. Extensive experiments show that MolMem achieves 90% success on single-property tasks (1.5\times over the best baseline) and 52% on multi-property tasks using only 500 oracle calls. Our code is available at https://github.com/REAL-Lab-NU/MolMem.

Lee Jung-Mok, Kim Sung-Bin, Joohyun Chang, Lee Hyun, Tae-Hyun Oh

Laughter is a complex social signal that conveys communicative intent beyond amusement. While prior work has focused on isolated laughter analysis tasks, a comprehensive understanding of laughter in real-world scenarios remains underexplored. We introduce SMILE-Next, a dataset for real-world laughter understanding with multimodal textual representations and question–answer annotations across three tasks: laughter detection, laughter type classification, and laughter reasoning. Building on this dataset, we propose a laughter expert LLM that leverages disentangled multimodal textual cues, together with a Mixture-of-Laugh-Experts framework and laughter-specific self-instruction for task-adaptive specialization. Experimental results show that the combination of our proposed components substantially outperforms multimodal LLM baselines, advancing robust real-world laughter understanding.

Faisal Hossain Raquib, Akm Moshiur Rahman Mazumder, Md Fahim, Md Tahmid Hasan Fuad, Md Farhan Ishmam, Faria Sultana, M Ashraful Amin, Amin Ahsan Ali, Akmmahbubur Rahman

Online safety in low-resource languages hinges not only on accurate hate speech detection but also on transparent, culturally grounded explanations. Yet prior works in Bangla largely focus on hate classification, while overlooking interpretability. We address this gap by introducing BanHADEX, the first hate explainability dataset in Bangla with human-annotated labels. BanHADEX contains 19,203 YouTube comments spanning April 2024–June 2025, annotated for binary hate classification with seven fine-grained hate categories, seven target groups, and concise explanations for each sample. Our data pipeline relies on a two-stage annotation protocol that uses majority voting for robust labeling. Our rich suite of experiments on open and closed-source LLMs reveals that explanation-guided LoRA substantially outperforms both classification and explanation quality across prompting and fine-tuning strategies. BanHADEX establishes the groundworks for faithful interpretability and safer moderation in linguistically rich yet under-resourced languages.

Zixuan Wang, Yuanyuan Lei

Logical reasoning serves as a central capability in LLMs and includes three main forms: deductive, inductive, and abductive reasoning. In this work, we study the knowledge representations of these reasoning types in LLMs and analyze the correlations among them. Our analysis shows that each form of logical reasoning can be captured as a reasoning-specific knowledge vector in a linear representation space, yet these vectors are largely independent of each other. Motivated by cognitive science theory that these subforms of logical reasoning interact closely in the human brain, as well as our observation that the reasoning process for one type can benefit from the reasoning chain produced by another, we further propose to refine the knowledge representations of each reasoning type in LLMs to encourage complementarity between them. To this end, we design a complementary subspace-constrained refinement framework, which introduces a complementary loss that enables each reasoning vector to leverage auxiliary knowledge from the others, and a subspace constraint loss that prevents erasion of their unique characteristics. Through steering experiments along reasoning vectors, we find that refined vectors incorporating complementary knowledge yield consistent performance gains. We also conduct a mechanism-interpretability analysis of each reasoning vector, revealing insights into the shared and specific features of different reasoning in LLMs.

Jin Cui, Jiaqi Guo, Jiepeng Zhou, Ruixuan Yang, Jiayi Lu, Jiajun Xu, Jiangcheng Song, Boran Zhao, Pengju Ren

While Large Language Models (LLMs) have emerged with remarkable capabilities in complex tasks through Chain-of-Thought (CoT) reasoning, practical resource constraints have sparked interest in transferring these abilities to smaller models. However, achieving both domain performance and cross-domain generalization remains challenging. Existing approaches typically restrict students to following a single golden rationale and treat different reasoning paths independently. Due to distinct inductive biases and intrinsic preferences, alongside the student’s evolving capacity and reasoning preferences during training, a teacher’s "optimal" rationale could act as out-of-distribution noise. This misalignment leads to a degeneration of the student’s latent reasoning distribution, causing suboptimal performance. To bridge this gap, we propose MIND, a capability-adaptive framework that transitions distillation from passive mimicry to active cognitive construction. We synthesize diverse teacher perspectives through a novel "Teaching Assistant" network. By employing a novel Feedback-Driven Inertia Calibration mechanism, this network utilizes inertia-filtered training loss to align supervision with the student’s current adaptability, effectively enhancing performance while mitigating catastrophic forgetting. Extensive experiments demonstrate that MIND achieves state-of-the-art performance on both in-distribution and out-of-distribution benchmarks, and our sophisticated latent space analysis further confirms the mechanism of reasoning ability internalization.