论文检索

Jiayuan Guo, Yueyang Su, Yuyao Ge, Saiping Guan, Lei Yu, Jiafeng Guo, Xueqi Cheng

Long context large language models exhibit the “lost in the middle” problem, where models struggle to effectively utilize information located in the middle of long contexts. Although existing workflow-based long context methods (e.g., RAG) alleviate this problem and perform well on specific datasets, can their effectiveness generalize to all types of datasets? In this work, we systematically investigate the cross-dataset generalization of long context methods. Our evaluation reveals that these methods are not universally effective. Such substantial performance variability underscores the risks of performance degradation associated with the indiscriminate application of long context methods. We investigated the reason for the failure of long context methods. We found that the intrinsic decomposition mechanisms of long context methods hinder context dependency modeling, causing these methods to suffer performance declines on documents with strong context dependency. To address this issue, We propose CoDaR (**Co**ntext **D**ependency-**a**ware **R**outing), a training-free adaptive routing strategy. By analyzing the context dependency strength of documents, CoDaR adaptively invokes long context methods, thereby significantly enhancing their overall robustness across different types of datasets.

Eunike Andriani Kardinata, Yusuke Sakai, Taro Watanabe

Acceptability judgments provide a crucial basis for understanding how sentences are perceived as natural or well-formed, and they are increasingly used to assess the linguistic capability of large language models (LLMs). Unlike grammaticality, acceptability depends not only on structural form but also on contextual and domain-specific factors. Most prior work evaluates sentences in isolation, and relatively little is known about how explicit contextual cues influence LLM acceptability judgments across domains. This study examines how contextual information affects model-generated acceptability ratings across multiple domains and several LLMs, using different forms of domain-specific contextual cues to situate sentences in their intended usage settings. The results show that context can meaningfully shift model judgments, although its effects vary across models and domains. Overall, the findings provide evidence on contextual effects in LLM acceptability judgment and support the development of more context-aware evaluation frameworks.

Hao Pang, Changcheng Li, Yingxue Liu

Language evolution is cognitively motivated by the reduction of communicative effort. Current research exploring this reported tendency has been constrained by the heavy reliance on manually annotated resources (e.g., dependency parsing) as well as a narrow focus (e.g., syntax as the single metric). To transcend these limitations, we propose two measures: Attention-based Structural Distance (ASD) and Semantic Space Distance (SSD). ASD is a parser-free measure of syntactic locality derived from the attention mechanism of pretrained large language models (LLM), while SSD is a measure of lexical distances that quantify the degree of separation between different parts of speech in the word vector space. Based on multiple diachronic and multilingual corpora, our experiments show a significant decrease of ASD while an increase of SSD, which implies a language developmental trend towards structural compactness and semantic divergence. Our research pioneers a novel lens grounded in LLM for studying language evolution, which has two major contributions. Linguistically, our study corroborates the hypothesized law of human language evolution by demonstrating that its development optimizes syntactic locality as well as functional semantic discriminability. Cognitively, our study shows that human and LLMs share common characteristics in language processing, lending support to the potential of employing LLMs in the study of human cognition.

Ijun Jang, Jewon Yeom, Juan Yeo, Hyunggyu Lim, Taesup Kim

Knowledge distillation (KD) is a widely adopted technique for transferring capabilities from large language models to smaller student models. However, conventional supervised KD often suffers from a distribution mismatch between training and inference. While on-policy KD approaches attempt to mitigate this issue by learning directly from student-generated outputs, they frequently encounter training instabilities and noisy teacher feedback during early optimization stages. These challenges manifest as pathological gradients in forward KL objectives when students encounter unfamiliar tokens, or as a collapse in distributional diversity within reverse KL regimes. To address these limitations, we propose Veto, an objective-level reformulation that constructs a geometric target distribution in logit space to emphasize agreement between the teacher and the student. By introducing a tunable parameter \beta, Veto serves as an Adaptive Gradient Veto that stabilizes optimization by suppressing harmful gradients on low-confidence tokens, while simultaneously acting as a Decisiveness Knob to balance reward-driven performance with output diversity. Extensive experiments across various reasoning and generation tasks demonstrate that Veto consistently outperforms supervised fine-tuning and existing on-policy baselines.

Aakriti Agrawal, Mucong Ding, Chenghao Deng, Zora Che, Arjun Rajaram, Anirudh Satheesh, Bang An, C. Bayan Bruss, John Langford, Furong Huang

With Large Language Models (LLMs) rapidly approaching and potentially surpassing human-level performance, it has become imperative to develop approaches capable of effectively supervising and enhancing these powerful models using smaller, human-level models exposed to only human-level data. We address this critical weak-to-strong (W2S) generalization challenge by proposing a novel method aimed at improving weak experts, by training on the same limited human-level data, enabling them to generalize to complex, super-human-level tasks. Our approach, called EnsemW2S, employs a token-level ensemble strategy that iteratively combines multiple weak experts, systematically addressing the shortcomings identified in preceding iterations. By continuously refining these weak models, we significantly enhance their collective ability to supervise stronger student models. We extensively evaluate the generalization performance of both the ensemble of weak experts and the subsequent strong student model across in-distribution (ID) and out-of-distribution (OOD) datasets. For OOD, we specifically introduce question difficulty as an additional dimension for defining distributional shifts. Our empirical results demonstrate notable improvements, achieving 4%, and 3.2% improvements on ID datasets and, upto 6% and 2.28% on OOD datasets for experts and student models respectively, underscoring the effectiveness of our proposed method in advancing W2S generalization.

Sagnik Anupam, Alexander Shypula, Osbert Bastani

Recent work has demonstrated the potential of large language models (LLMs) for program optimization, a key challenge in programming languages. We propose a blackbox adaptation method called Retrieval Augmented Search (RAS) that performs beam search over candidate optimizations; at each step, it retrieves in-context examples from a given training dataset of slow-fast program pairs to guide the LLM. Critically, we find that performing contextual retrieval based on an LLM-generated natural language description significantly outperforms retrieval based on the source code. We also propose AEGIS, a method for improving interpretability by decomposing training examples into "atomic edits” that are significantly more incremental in nature. We show that RAS performs up to 2.06\times better than prior state-of-the-art blackbox adaptation strategies on optimizing C++ programs, and that AEGIS performs up to 1.37\times better while making significantly smaller edits. We also show that using RAS improves the mean runtime percentile of Python programs by 10.27 compared to baselines.

Weixiang Zhao, Yichen Zhang, Yingshuo Wang, Yang Deng, Yanyan Zhao, Xuda Zhi, Yongbo Huang, Hao He, Wanxiang Che, Bing Qin 等

Experience-driven self-evolution has emerged as a promising paradigm for improving the autonomy of large language model agents, yet its reliance on self-curated experience introduces underexplored safety risks. In this study, we investigate how experience accumulation and utilization in self-evolving agents affect safety performance across web-based and embodied environments. Notably, experience gathered solely from benign tasks can still compromise safety in high-risk scenarios. Further analysis attributes this degradation to the execution-oriented nature of accumulated experience, which reinforces agents’ tendency to act rather than refuse. In more realistic settings where agents encounter both benign and harmful tasks, refusal-related experience mitigates safety decline but induces over-refusal, revealing a fundamental safety–utility trade-off. Overall, our findings expose inherent limitations of current self-evolving agents and call for more principled strategies to ensure safe and reliable adaptation.

Wei Fang, James R. Glass

LLM agents operating over massive, dynamic tool libraries rely on effective retrieval, yet standard single-shot dense retrievers struggle with complex requests. These failures primarily stem from the disconnect between abstract user goals and technical documentation, and the limited capacity of fixed-size embeddings to model combinatorial tool compositions. To address these challenges, we propose ToolQP, a lightweight framework that models retrieval as iterative query planning. Instead of single-shot matching, ToolQP decomposes instructions into sub-tasks and dynamically generates queries to interact with the retriever, effectively bridging the semantic gap by targeting the specific sub-tasks required for composition. We train ToolQP using synthetic query trajectories followed by optimization with Reinforcement Learning with Verifiable Rewards (RLVR). Experiments demonstrate that ToolQP achieves state-of-the-art performance, exhibiting superior zero-shot generalization, robustness across diverse retrievers, and significant improvements in downstream agentic execution.

Pengyuan Qin, Linnan Tu, Yuhan Ke, Hefei Ling

Large language models (LLMs) watermarking has been proposed as an active approach for content provenance verification, yet existing evaluations are largely confined to fixed entropy settings. In this paper, we introduce EntroBench, a benchmark for LLM watermarking that systematically covers three entropy levels and seven representative tasks. We conducted a fair evaluation of eight watermarking methods through hyper-parameter search based on an anchored dataset. We find that current approaches struggle to perform consistently across different entropy levels. Our analysis reveals a clear trade-off between watermark detectability and downstream output quality that varies across tasks and entropy conditions. Furthermore, we assess watermark robustness under realistic user interaction scenarios and show that common, non-adversarial user behaviors can substantially degrade watermark signals. These results indicate that practical usage-driven perturbations pose a significant challenge to current watermarking techniques. EntroBench provides a unified evaluation framework for studying these issues and supports the development of more adaptive and robust LLM watermarking methods. Dataset and codes are available at https://github.com/py-qin/EntroBench.

Hong Kyu Lee, Ruixuan Liu, Li Xiong

Machine unlearning is an emerging technique that removes the influence of a subset of training data (forget set) from a model without full retraining, with applications including privacy protection, content moderation, and model correction. The key challenge lies in achieving strong unlearning efficacy while preserving the overall utility. Existing unlearning methods for large language models (LLMs) often rely on auxiliary models, retain datasets, or even commercial AI services. However, dependence on these external resources is often impractical and could potentially introduce additional privacy risks. In this work, we propose direct token optimization (DTO), a self-contained unlearning approach for LLMs that directly optimizes the token-level objectives to unlearn specific sequences without external resources.For each sequence to be unlearned, we identify target tokens that encode critical knowledge for unlearning and treat remaining tokens as non-target ones for maintaining the model utility. DTO maximizes an unlearning objective on target tokens and applies a utility-preservation regularizer on non-target tokens.Across multiple unlearning benchmarks, DTO improves the forget quality up to 16.8\times over the latest baselines while maintaining comparable model utility. Our code is available at github.com/Emory-AIMS/direct_token_optimization.

Pardis Sadat Zahraei, Xiaoning Wang, Nimet Beyza Bozdag, Gokhan Tur, Dilek Hakkani-Tür

Large Language Models (LLMs) are increasingly used as judges to evaluate text quality, moderate content, and assess arguments. We investigate whether alignment-instilled prior beliefs bias LLM judgments, using persuasion evaluation as a representative task. We find a systematic failure: models conflate their trained beliefs with rhetorical quality, rating identical claims differently based on belief alignment rather than argumentative merit. A bare assertion aligned with training receives higher scores than a well-crafted counter-argument, even when explicitly instructed to judge rhetoric alone. We introduce ConvinceQA, a dataset of 27,756 persuasive arguments with controlled stance variation across subjective, harmful, and misinformation domains, and demonstrate this prior prejudice across models. We exploit this failure through persuasion-based probing: evaluating minimal pairs that differ only in the subject token bypasses learned refusals and reveals hidden biases. Analysis identifies three failure modes, with belief-conditioned rating inflation accounting for 88% of cases. Cross-task validation on essay quality assessment and debate judging confirms this is a pervasive limitation.

Aakriti Agrawal, Gouthaman KV, Rohith Aralikatti, Gauri Jagatap, Jiaxin Yuan, Sarvesh Baskar, Vijay Kamarshi, Andrea Fanelli, Furong Huang

Hallucinations in Large Vision-Language Models (LVLMs) remain a persistent challenge, often stemming from inadequate integration of visual information during multimodal reasoning. A key cause is the model’s over-reliance on textual priors and underutilization of visual cues, leading to outputs that are linguistically fluent but visually inaccurate. For example, given an image of an empty kitchen countertop, an LVLM might hallucinate a “bowl of fruit” or “cup of coffee,” relying on language associations rather than visual evidence. Most LVLMs incorporate visual features by appending them to the input stream of a pre-trained LLM and training on large-scale vision-language datasets. Our systematic analysis reveals that this strategy often leads to over-dependence on textual information due to the inherent bias of LLMs towards language-dominant representations. This imbalance skews attention towards the text over visual content, weakening the model’s ability to ground outputs in visual inputs. To address this, we propose a simple yet effective visual feature incorporation method that encourages the model to learn visually-informed textual embeddings distinct from those of the base LLM and promotes a more balanced attention distribution. Experimental results across multiple hallucination benchmarks demonstrate that our method significantly reduces hallucinations and fosters more balanced multimodal reasoning. Notably, our approach achieves substantial gains, including \textbf{+9.33}% on MMVP-MLLM, \textbf{+2.99}% on POPE-AOKVQA, up to \textbf{+3.4}% on Merlin, and \textbf{+3}% on the hard-data split of HallusionBench.

Wonje Jeung, Sangyeon Yoon, Hyesoo Hong, Soeun Kim, Seungju Han, Youngjae Yu, Albert No

Machine unlearning aims to remove “forget” data while preserving knowledge from the “retain” data, yet a fundamental question arises when the two share content. By definition, an unlearned model should be indistinguishable from a model retrained solely on the retain set, which implies that shared knowledge must remain while only forget-specific content is removed. To evaluate this requirement, we introduce DUSK, the first benchmark for unlearning under realistic knowledge overlap. DUSK constructs documents containing both shared and unique knowledge and defines seven metrics to test whether methods erase forget-specific expressions without discarding shared facts. Evaluating nine recent approaches, we find that although surface text is often removed, current methods struggle to distinguish shared from unique knowledge, either erasing information that should be retained or failing to fully forget target content. DUSK provides a controlled, reproducible testbed for diagnosing these failures and guiding precise unlearning algorithms.

Qipeng Xie, Zi Liang, Jiafei Wu, Yufei Chen, Weizheng Wang, Wenao Ma, Zhong Ming, Haiqin Yang, Kaishun Wu

Large Language Models (LLMs) often exhibit extreme sensitivity to surface-level prompt variations, where minor lexical perturbations trigger disproportionate performance fluctuations. Moving beyond black-box optimization or coarse-grained templates, we conduct the first analysis of n-gram token-level mechanisms, leveraging a large-scale dataset of 132,000 prompt variants. Our investigation uncovers the Scaling Law of Prompt Performance Stability: higher average performance is inherently associated with lower variance and greater stability. We identify that this robustness is driven by two linguistic pillars: Domain-Specific Terminology, which anchors semantic boundaries, and Explicit Action Directives, which formalize reasoning trajectories. By narrowing the model’s interpretative space, these patterns effectively "lock" the generation process. We operationalize these findings into an automated Prompt-Refining Agent that autonomously restructures queries via domain anchoring and operational constraints. Empirical results show a 40.7% reduction in performance variance for code generation, offering a statistically grounded framework for robust prompt engineering.

Prerna Agarwal, Ayushman Kumar Singh, Srikanta Bedathur

Existing low-resource in-context learning-based knowledge graph question answering (KGQA) methods rely heavily on large language models (LLMs) to convert the natural language question into its corresponding logical form (LF), such as SPARQL, KoPL, etc. Recently, a few alignment techniques have been introduced that enable instruction-based fine-tuning of language models. They provide explicit negative signals and comparative objectives to learn how to avoid negative signals using preference optimization methods. Exploring such fine-tuning techniques with LLMs becomes very challenging due to the high computational resource requirements associated with them. Due to this, the focus has been shifted towards Small Language Models (SLMs), which offer advantages such as ease of (i) deployment for practical applications and (ii) instruction fine-tuning for specialized tasks. Motivated by this, in this work, we propose PO-KGQA: An SLM-based preference optimization framework for the complex KGQA task in a low-resource setting. Our extensive experiments demonstrate how PO-KGQA outperforms other fine-tuning alignment techniques on complex benchmarks such as KQA Pro by approximately 9% (avg).

Kirandeep Kaur, Vinayak Gupta, Aditya Gupta, Chirag Shah

Current approaches to proactive assistance move beyond the ask-and-respond paradigm by anticipating user needs. In practice, they either burden users with clarifying questions or rely on context-based extrapolation, often leading to unnecessary or mistimed interventions. Such systems lack explicit mechanisms to model users’ knowledge gaps, resulting in incomplete or suboptimal task outcomes. To address this, we propose PROPER, a framework that explicitly models user-specific knowledge gaps in a controlled manner. Central to our approach is the notion of dimensions: structured, task-relevant factors that define the considerations required for effective task completion. Given a user query, the DGA (Dimension Generating Agent) identifies explicit dimensions (from the user’s query) and generates a set of candidate implicit dimensions capturing unarticulated aspects of the task. The RGA (Response Generating Agent) integrates both explicit and implicit dimensions selectively to produce personalized, context-aware, and proactively informative responses. We evaluate PROPER across multiple domains using a structured, gap-aware rubric that measures coverage, initiative appropriateness, and intent alignment. PROPER improves on quality scores and win rates across all domains, achieving up to 84% gains in single-turn evaluation and consistent dominance in multiturn interactions. All code for PROPER is available at: https://github.com/i-kiran/ProPer-Agent.

Aryan Roy, Zekun Wang, Christopher J. MacLellan

Do vision-language models (VLMs) develop more human-like sensitivity to linguistic concreteness than text-only large language models (LLMs) when both are evaluated with text-only prompts? We study this question with a controlled comparison between matched Llama text backbones and their Llama Vision counterparts across multiple model scales, treating multimodal pretraining as an ablation on perceptual grounding rather than access to images at inference. We measure concreteness effects at three complementary levels: (i) output behavior, by relating question-level concreteness to QA accuracy; (ii) embedding geometry, by testing whether representations organize along a concreteness axis; and (iii) attention dynamics, by quantifying context reliance via attention-entropy measures. In addition, we elicit token-level concreteness ratings from models and evaluate alignment to human norm distributions, testing whether multimodal training yields more human-consistent judgments. Across benchmarks and scales, VLMs show larger gains on more concrete inputs, exhibit clearer concreteness-structured representations, produce ratings that better match human norms, and display systematically different attention patterns consistent with increased grounding.

Xing Fu, Yulin Hu, Mengtong Ji, Haozhen Li, Yixin Sun, Weixiang Zhao, Yanyan Zhao, Bing Qin

Memory-augmented language agents are increasingly deployed in affective applications such as emotional support, where understanding and responding to users’ latent emotional needs is critical. However, existing research often treats memory as a tool for factual retrieval, overlooking its role in shaping users’ emotional experiences. In this work, we introduce ENPMR-Bench, a benchmark for evaluating Emotional Need-aware Proactive Memory Retrieval (ENPMR), a core capability that enables agents to infer users’ latent emotional needs and proactively retrieve appropriate memories to support empathetic interaction. Grounded in Maslow’s hierarchy of needs, ENPMR-Bench includes over 1,800 memory-augmented dialogues and defines structured mappings between emotional needs and supportive memory types. Experimental results demonstrate that current retrieval paradigms, including both embedding-based and LLM-driven approaches, exhibit substantial deficiencies, with empathy scores significantly lagging behind golden memory conditions. While chain-of-thought prompting improves the alignment between inferred emotional needs and retrieved memories to some extent, a notable performance gap remains. Together, these findings reveal critical limitations in current agents and outline directions for advancing personalized emotional support through need-sensitive memory retrieval.

Zhining Liu, Tianyi Wang, Xiao Lin, Penghao Ouyang, Gaotang Li, Ze Yang, Hui Liu, Sumit Keswani, Vishwa Pardeshi, Huijun Zhao 等

Despite substantial efforts toward improving the moral alignment of Vision-Language Models (VLMs), it remains unclear whether their ethical judgments are stable in realistic settings. This work studies moral robustness in VLMs, defined as the ability to preserve moral judgments under textual and visual perturbations that do not alter the underlying moral context. We systematically probe VLMs with a diverse set of model-agnostic multimodal perturbations and find that their moral stances are highly fragile, frequently flipping under simple manipulations. Our analysis reveals systematic vulnerabilities across perturbation types, moral domains, and model scales, including a sycophancy trade-off where stronger instruction-following models are more susceptible to persuasion. We further show that lightweight inference-time interventions can partially restore moral stability. These results demonstrate that moral alignment alone is insufficient and that moral robustness is a necessary criterion for the responsible deployment of VLMs.

Pratham Singla, Shivank Garg, Ayush Singh, Ishan Garg, Ketan Suhaas Saichandran

Recent advances in post-training techniques have endowed Large Language Models (LLMs) with enhanced capabilities for tackling complex, logic-intensive tasks through the generation of supplementary planning tokens. This development raises a fundamental question – Are these models aware of what they "learn” and "think”? To address this, we define three core competencies: (1) awareness of learned latent policies, (2) generalization of these policies across domains, and (3) alignment between internal reasoning traces and final outputs. We empirically evaluate these abilities on several tasks, each designed to require learning a distinct policy. Furthermore, we contrast the profiles of models post-trained via Supervised Fine-Tuning (SFT), Direct Policy Optimization (DPO), and Group Relative Policy Optimization (GRPO). Our findings indicate that RL-trained models not only demonstrate greater awareness of their learned behaviors and stronger generalizability to novel, structurally similar tasks than SFT models but also often exhibit weak alignment between their reasoning traces and final outputs, an effect most pronounced in GRPO-trained models.