论文检索

Canran Wang, Yuwen Yang, Zhen Wang, Ming MA, Ding Yu, Chentai Wang, Keman Huang, Xiaoyong Du

The double-edged sword of integrating Large Language Models (LLMs) requires an effective triadic collaboration mechanism among LLMs, teachers and students, especially for K-12 education. By developing a triadic collaboration system to support K-12 writing learning, a multidimensional evaluation framework grounded in Systemic Functional Linguistics and the suggestion trajectory tracing pipeline, this paper contributes a large-scale empirical dataset involving 57,954 essays from 10,195 students across 120 schools over two years. Our findings confirm the efficacy of this system in improving writing quality through a strategic labor division: the LLM serves as a generative engine to mitigate teacher burnout, and the teacher acts as a pedagogical gatekeeper and bridge to guarantee feedback quality. While both LLM and teacher are critical for skill improvement, we uncover a ceiling effect where excessive linguistic expansion yields diminishing marginal utility. These suggest a dynamically adaptive LLM-teacher collaboration as student proficiency increases.

Yaocheng Zhang, Haohuan Huang, Zijun Song, Zijie Zhao, Qichao Zhang, Yuanheng Zhu, Dongbin Zhao

Tool-Integrated Reasoning (TIR) with search engines enables large language models to iteratively retrieve up-to-date external knowledge, enhancing adaptability and generalization in complex question-answering tasks. However, existing search agent pipelines typically depend on reinforcement learning based optimization, which often suffers from sparse outcome rewards, leading to inefficient exploration and unstable training. We introduce CriticSearch, a fine-grained credit-assignment framework that supplies dense, turn-level feedback via a retrospective critic mechanism. During training, a frozen, asymmetric critique LLM retrospectively evaluates each turn using privileged information from the full trajectory and gold answers, converting these assessments into stable, dense rewards that guide policy improvement. Experimental results across diverse multi-hop reasoning benchmarks demonstrate that CriticSearch consistently outperforms existing baselines, achieving faster convergence, improved training stability, and higher performance.

Shannon Zejiang Shen, Valerie Chen, Ken Gu, Alexis Ross, Zixian Ma, Jillian Ross, Alex Gu, Chenglei Si, Wayne Chi, Andi Peng 等

Current evaluations of agents remain centered around one-shot task completion, failing to account for the inherently iterative and collaborative nature of many real-world problems, where human goals are often underspecified and evolve. We argue for a shift from building and assessing task completion agents to developing collaborative agents, assessed not only by the quality of their final outputs but by how well they engage with and enhance human effort throughout the problem-solving process. To support this shift, we introduce collaborative effort scaling, a framework that captures how an agent’s utility grows with increasing user involvement. Through case studies and simulated evaluations, we show that state-of-the-art agents often underperform in multi-turn, real-world scenarios, revealing a missing ingredient in agent design: the ability to sustain engagement and scaffold user understanding. Collaborative effort scaling offers a lens for diagnosing agent behavior and guiding development toward more effective interactions.

Yuuki Tachioka

Large language models (LLMs) exhibit non-stationary generation: their output distributions shift with prompts, retrieved documents, and decoding conditions. Under such variability, average likelihood metrics can obscure heterogeneous behaviors across samples, especially in high-surprisal tails where failures often occur. We propose an information-spectrum-based diagnostic framework that treats LLMs as general sources without assuming stationarity, ergodicity, or the asymptotic equipartition property. We define sequence-level self-information density (coding rate; mean surprisal) and construct an empirical information spectrum from finite samples, enabling operational estimates of spectrum quantiles and width. We further introduce an information gain spectrum, a teacher-forced likelihood-based measure that evaluates the same generated sequence with and without side information. Across multiple Japanese LLMs and QA settings, we observe that correctness differences are often more visible in the high-surprisal tail than in the mean coding rate, and that side information can reshape tail behavior in heterogeneous ways across sequences. We also observe that instruction tuning changes the spectrum structure, making tail statistics and spectrum width more predictive of correctness than the mean coding rate. Overall, our analysis illustrates how spectrum-based diagnostics complement average-based metrics for understanding conditional generation.

Saman Sarker Joy, Swakkhar Shatabda

Large-scale multitask benchmarks have driven rapid progress in language modeling, yet most emphasize high-resource languages such as English, leaving Bengali underrepresented. We present BnMMLU, a comprehensive benchmark for measuring massive multitask language understanding in Bengali. BnMMLU spans 41 domains across STEM, humanities, social sciences, and general knowledge, and contains 134,375 multiple-choice question–option pairs-the most extensive Bengali evaluation suite to date. The dataset preserves mathematical content via MathML, and includes BnMMLU-HARD, a compact subset constructed from questions most frequently missed by top systems to stress difficult cases. We benchmark 24 model variants across 11 LLM families, spanning open-weights general/multilingual, Bengali-centric open-weights, and proprietary models, covering multiple parameter scales and instruction-tuned settings. We evaluate models under standardized protocols covering two prompting styles (Direct vs. Chain-of-Thought) and two context regimes (0-shot vs. 5-shot), reporting accuracy consistently across families. Our analysis highlights persistent gaps in reasoning and application skills and indicates sublinear returns to scale across model sizes. We release the dataset and evaluation templates to support rigorous, reproducible assessment of Bengali language understanding and to catalyze progress in multilingual NLP.

Yuting Huang, Yiquan Wu, Meitong Guo, Ang Li, Xiaozhong Liu, Keting Yin, Fei Wu, Kun Kuang

Artificial intelligence has become increasingly prevalent in the legal domain. However, LegalAI systems often struggle with vague user queries that lack essential legal details, leading to suboptimal performance in practical applications. To address this challenge, we propose FactFiller, a novel approach that dynamically generates questionnaires to help users refine their input queries. Our method leverages an iterative training process that collects valuable questionnaires, eliminating the need for human annotation. Additionally, we introduce a "case-law-quiz” cascading retrieval process, ensuring that the generated questions and answer options are directly linked to specific legal provisions. Through the user study and the downstream task experiments, we demonstrate that FactFiller, while remaining easy for non-experts to understand, not only improves the completeness of queries but also ensures the performance of various domain-specific models in downstream legal tasks.

Jessica H Zhu, Shayla Stringfield, Vahe Zaprosyan, Michael Wagner, Michel Cukier, Joseph Richardson

Firearm violence is a pressing public health issue, yet research into survivors’ lived experiences remains underfunded and difficult to scale. Qualitative research, including in-depth interviews, is a valuable tool for understanding the personal and societal consequences of community firearm violence and designing effective interventions. However, manually analyzing these narratives through thematic analysis and inductive coding is time-consuming and labor-intensive. Recent advancements in large language models (LLMs) have opened the door to automating this process, though concerns remain about whether these models can accurately and ethically capture the experiences of vulnerable populations. In this study, we assess the use of open-source LLMs to inductively code interviews with 21 Black men who have survived community firearm violence. Our results demonstrate that while some configurations of LLMs can identify important codes, overall relevance remains low and is highly sensitive to data processing. Furthermore, LLM guardrails lead to substantial narrative erasure. These findings highlight both the potential and limitations of LLM-assisted qualitative coding and underscore the ethical challenges of applying AI in research involving marginalized communities.

Kaize Shi, Xueyao Sun, Xiaohui Tao, Lin Li, Qika Lin, Guandong Xu

Large Language Models (LLMs) face information overload when handling long contexts, particularly in Retrieval-Augmented Generation (RAG) where extensive supporting documents introduce redundant content that interferes with reasoning. Context engineering has emerged to address these challenges, yet existing methods rely on lexical or token-level features that fragment semantic units and fail to capture conceptually essential content. We propose an unsupervised context compression framework leveraging Abstract Meaning Representation (AMR) to preserve semantically essential information while filtering irrelevant text. By quantifying node-level entropy within AMR graphs, our method estimates the conceptual importance of each node, enabling retention of core semantics. Specifically, we construct AMR graphs from retrieved contexts, compute the conceptual entropy of each node, and identify statistically significant concepts to form a condensed, semantically focused context. Experiments on the PopQA and EntityQuestions datasets demonstrate that our method outperforms vanilla RAG and existing baselines, achieving superior accuracy while substantially reducing context length. To the best of our knowledge, this is the first work introducing AMR-based conceptual entropy for context compression, demonstrating the potential of structured linguistic representations in context engineering.

Shangbin Feng, Yike Wang, Weijia Shi, Yulia Tsvetkov

We propose Data Swarms, an algorithm to optimize the generation of synthetic evaluation data and advance quantitative desiderata of LLM evaluation. We first train a swarm of initial data generators using existing data, and define various evaluation objectives to reflect the desired properties of evaluation (e.g., generate more difficult problems for the evaluated models) and quantitatively evaluate data generators. We then employ particle swarm optimization to optimize the swarm of data generators, where they collaboratively search through the model parameter space to find new generators that advance these objectives. We further extend it to Adversarial Swarms, where the data generator swarm generates harder data while the test taker model swarm learns from such data, co-evolving dynamically for better data and models simultaneously. Extensive experiments demonstrate that Data Swarms outperforms eight data generation baselines across five evaluation objectives, while Adversarial Swarms produce more robust learning of synthetic data and stronger generalization. Further analysis reveals that Data Swarms successfully optimizes compositions of multiple evaluation objectives and generalizes to new off-the-shelf LLMs, unseen at optimization time.

Hanqi Yan, Xiangxiang Cui, Lu Yin, Jindong Gu, Paul Pu Liang, Yulan He, Yifei Wang

The success of vision-language models is primarily attributed to effective cross-modal alignment between vision and language. However, modality gaps persist even in well-aligned models and may be necessary for human perception, as evidenced by modality-specific phenomena such as visual texture and linguistic tone. These observations motivate us to computationally measure and leverage modality gaps to explore their utility in downstream applications. In this paper, we introduce the Modality Dominance Score (MDS), which attributes multimodal features to specific modalities by categorizing them as vision-dominant, language-dominant, or cross-modal. We then propose automatic interpretability metrics to evaluate these modality-specific features in a scalable manner. Finally, we demonstrate how the identified modality-specific features enable training-free probing and editing methods for understanding model perception across genders, generating adversarial examples, and controlling text-to-image generation. Combined with task-agnostic interpretability tools, our work provides a systematic framework for analyzing and efficiently controlling multimodal models.

Wei-Rui Chen, Vignesh Kothapalli, Ata Fatahibaarzi, Hejian Sang, Shao Tang, Qingquan Song, Zhipeng Wang, Muhammad Abdul-Mageed

Distilling the capabilities from a large reasoning model (LRM) to a smaller student model often involves training on substantial amounts of reasoning data. However, knowledge distillation (KD) over lengthy sequences with prompt (P), chain-of-thought (CoT), and answer (A) sections makes the process computationally expensive. In this work, we investigate how the allocation of supervision across different sections (P, CoT, A) affects student performance. Our analysis shows that selective KD over only the CoT tokens can be effective when the prompt and answer information is encompassed by it. Building on this insight, we establish a truncation protocol to quantify computation-quality tradeoffs as a function of sequence length. We observe that beyond a specific length, longer training sequences provide marginal returns for downstream performance but require substantially higher memory and FLOPs. To this end, training on only the first 50% of tokens of every training sequence can retain, on average, ≈91% of full-sequence performance on math benchmarks while reducing training time, memory usage, and FLOPs by about 50% each. Codes are available at https://github.com/weiruichen01/distilling-the-essence.

Yizhuo Zhang, Heng Wang, Shangbin Feng, Zhaoxuan Tan, Xinyun Liu, Yulia Tsvetkov

Previous research has sought to enhance the graph reasoning capabilities of LLMs by supervised fine-tuning on synthetic graph data. While these led to specialized LLMs better at solving graph algorithm problems, we don’t need LLMs for shortest path: we need generalization from synthetic graph data to real-world tasks with implicit graph structures. In this work, we propose to unlock generalizable learning of graph with post-training alignment with synthetic data. We first design solution-based and process-based rewards for synthetic graph problems: instead of rigid memorizing response patterns in direct fine-tuning, we posit that post-training alignment would help LLMs grasp the essentials underlying graph reasoning and alleviate overfitting on synthetic data. We employ post-training alignment algorithms such as GRPO and DPO, aligning both off-the-shelf LLMs and LLMs fine-tuned on synthetic graph data. We then compare them against existing settings on both in-domain synthetic tasks and out-of-domain real-world tasks with implicit graph structures such as multi-hop QA, structured planning, and more. Extensive experiments demonstrate that our post-training alignment recipe leads to statistically significant improvement on 5 datasets, with an average gain of 12.9% over baseline settings. Further analysis reveals that process-based rewards consistently outperform solution-based rewards on synthetic data but not on real-world tasks, and compositionality and explainable intermediate steps remains a critical challenge even after post-training alignment.

Saloni Dash, Amélie Reymond, Emma Spiro, Aylin Caliskan

Reasoning in humans is prone to biases due to underlying motivations like identity protection, that undermine rational decision-making and judgment. This motivated reasoning at a collective level can be detrimental to society when debating critical issues such as human-driven climate change or vaccine safety, and can further aggravate political polarization. Prior studies have reported that large language models (LLMs) are also susceptible to human-like cognitive biases, however, the extent to which LLMs selectively reason toward identity-congruent conclusions remains largely unexplored. Here, we investigate whether assigning 8 personas across 4 political and socio-demographic attributes induces motivated reasoning in LLMs. Testing 8 LLMs (open source and proprietary) across two reasoning tasks from human-subject studies — veracity discernment of misinformation headlines and evaluation of numeric scientific evidence — we find that persona-assigned LLMs have up to 9% reduced veracity discernment relative to models without personas. Political personas specifically are up to 90% more likely to correctly evaluate scientific evidence on gun control when the ground truth is congruent with their induced political identity. Prompt-based debiasing methods are largely ineffective at mitigating these effects. Taken together, our empirical findings are the first to suggest that persona-assigned LLMs exhibit human-like motivated reasoning that is hard to mitigate through conventional debiasing prompts — raising concerns of exacerbating identity-congruent reasoning in both LLMs and humans.

Ruozhen Yang, Yucheng Jiang, Yueqi Jiang, Priyanka Kargupta, Yunyi Zhang, Jiawei Han

Deploying large language models in long-horizon, goal-oriented interactions remains challenging because similar entities and facts recur under different latent goals and con-straints, causing memory systems to retrieve context-mismatched evidence. We propose STITCH (Structured Intent Tracking in Contextual History), an agentic memory system that indexes each trajectory step with a structured retrieval cue, contextual intent, and retrieves history by matching the current step’s intent. Contextual intent provides compact signals that disambiguate repeated mentions and reduce interference: (1) the current latent goal defining a thematic segment, (2) the action type, and (3) the salient entity types anchoring which attributes matter. During inference, STITCH filters and prioritizes memory snippets by intent compatibility, suppressing semantically similar but context-incompatible history.For evaluation, we introduce CAME-Bench, a benchmark for context-aware retrieval in realistic, dynamic, goal-oriented trajectories. Across CAME-Bench and LongMemEval, STITCH achieves state-of-the-art performance, outperforming the strongest baseline by 35.6%, with the largest gains as trajectory length increases. Our analysis shows that intent indexing substantially reduces retrieval noise, supporting intent-aware memory for robust long-horizon reasoning.

Christian Giannetti, Giovanni Trappolini, Nicola Tonellotto, Fabrizio Silvestri, Pietro Lio

Large language models exhibit a critical vulnerability to distractor interference in retrieval-augmented contexts: they fail to prioritize relevant, factually correct documents over topically similar but misleading content. We introduce Lat-Defuse, a mechanistic framework that corrects this failure mode through targeted interventions in the model’s latent space. Using Sparse Autoencoders (SAEs), our method operates in an interpretable feature space and formulates correction as constrained counterfactual optimization. On Gemma-2 and Llama-3 model families across three QA benchmarks (BioASQ, Natural Questions, PopQA), our method achieves recovery rates of up to 94% on distractor-vulnerable samples. Successful correction through sparse modifications reveals distractor interference as a localized, systematically addressable phenomenon, opening directions toward universal distractor robustness in LLMs.

Zonghao Ying, Yangguang Shao, Jianle Gan, Gan Xu, Wenxin Zhang, Quanchen Zou, Junzheng Shi, Zhenfei Yin, Mingchuan Zhang, Aishan Liu 等

Large vision–language model (LVLM)-based web agents are emerging as powerful automation tools but face severe security risks in real-world deployment. Existing benchmarks offer limited coverage, typically isolating user-level prompts from environmental threats, thus failing to capture the full spectrum of vulnerabilities. To address this, we present SecureWebArena, the first holistic security benchmark for web agents. SecureWebArena features a unified suite of six realistic web environments with 2,970 adversarial trajectories, covering a structured taxonomy of six attack vectors that span both user-level and environment-level manipulations. Crucially, we introduce a multi-layered evaluation protocol that dissects agent failures across internal reasoning, behavioral execution, and task outcomes, enabling fine-grained risk analysis beyond simple success metrics. Experiments on 9 representative LVLMs reveal universal vulnerabilities to subtle manipulations and uncover significant trade-offs between model specialization and security. SecureWebArena establishes a rigorous foundation for advancing the development of trustworthy web agents.

Weiming Li, Hye-young Paik, Yulei Sui

Multimodal GUI agents generally operate on raw visual and textual observations, which creates a fundamental scalability challenge. While current state-of-the-art frameworks predominantly rely on inference-intensive test-time scaling or the accumulation of unbounded raw logs to maintain task coherence, we attribute the underlying bottleneck to insufficient state abstraction.To address this, we propose HiSA, a hierarchical state abstraction approach that actively restructures knowledge rather than passively retaining historical information by organizing raw histories into a three-level hierarchy of abstracted steps, refined contexts, and induced patterns.By synthesizing high-dimensional observations into compact semantic states, HiSA decouples reasoning efficacy from context length, enabling precise and scalable decision-making as interaction histories grow.When evaluating using Spider2-V, our approach establishes a new state-of-the-art, achieving a 40.58% success rate while reducing token consumption by 69.85% and monetary costs by 55.10% compared to the best-performing baseline.

Ryuji Hashimoto, Masahiro Kaneko, Ryosuke Takata, Takehiro Takayanagi, Kiyoshi Izumi

Opinion dynamics (OD) studies how individual opinions evolve and generate collective patterns such as consensus and polarization. While recent work explores OD using populations of LLM-based agents focusing on opinion exchange, it typically does not incorporate individuals’ lived experiences, such as economic outcomes of past decisions, which play a critical role in shaping opinions. We propose a novel OD simulation framework that grounds LLM-based agents in an economic environment, allowing them to act and receive environmental feedback. Our simulations exhibit coherent OD at both individual and population levels: individual opinions follow structured trajectories shaped by economic experiences, with adverse conditions inducing opinion rigidity, while at the population level, collective opinions co-move with economic conditions, with inequality amplifying polarization and price instability driving larger distributional shifts. These results highlight the importance of grounding LLM-based agents in environments to capture collective OD.

Xinyuan Song, Ziyi Ni, Fred Yang, Bo Zhang, Yijin Wang, Jane Zhang

Automated assessment of patent quality is increasingly important given the growth of patent filings and the adoption of AI-assisted drafting. Existing methods often rely on modular pipelines or generic detectors, resulting in fragmented decisions and limited integration across quality dimensions. We propose P-QuASAR (Patent Quality Assurance via Structured Assessment and Refinement), a unified probabilistic framework that represents patent specifications as Quality Graphs. Multiple interdependent quality dimensions—such as regulatory compliance, technical coherence, and figure–text consistency—are jointly modeled using uncertainty-aware Quality Assessment Functions with learned edge potentials. Cross-dimensional evidence propagation via loopy belief propagation enables calibrated defect detection, while Optimal Intervention Paths translate inferred quality states into prioritized and actionable refinement recommendations. Evaluated on 500 patents across eight IPC domains against seven state-of-the-art baselines, P-QuASAR achieves substantial improvements: 99.86% balanced accuracy on regulatory compliance, 88.91% on technical coherence, and 94.70% on figure consistency, outperforming the strongest baselines by 3.0%, 9.0%, and 7.1%, respectively. Ablation studies confirm that joint graph reasoning contributes 3.66 points to average performance. When applied for refinement, P-QuASAR reduces average defects in AI-generated patents from 9.04–12.15 to 3.21 per document, surpassing human-authored patents.

Zijie Wang, MohammadHossein Rezaei, Farzana Rashid, Eduardo Blanco

Negation is a common and important semantic feature in natural language, yet Large Language Models (LLMs) struggle when negation is involved in natural language understanding tasks. Commonsense knowledge, on the other hand, despite being a well-studied topic, lacks investigations involving negation. In this work, we show that commonsense knowledge with negation is challenging for models to understand. We present a novel approach to automatically augment existing commonsense knowledge corpora with negation, yielding two new corpora containing over 2M triples with if-then relations. In addition, pre-training LLMs on our corpora benefits negation understanding.