跳到正文
研踪PaperTrail

论文库

检索计算机领域论文,按研究方向、作者或关键词筛选。

部分采集异常
正在读取账号与研究记录…

全站更新概况

+729新增论文
+575其中 AI 相关
+82已有论文修订
+95新增 AI 导读

arXiv 最近完成扫描:2026-11-21 12:19 · 全站最近入库:2026-11-21 12:19各信源进度 →

统计口径与数据新鲜度

统计 2026-11-20 15:13 — 2026-11-21 15:13(北京时间) · 点击数字查看对应简报;AI 相关包含在新增论文内。

最近完成检查:2026-11-21 12:19最近发现新增版本:2026-11-21 12:19库内论文最新更新时间:2026-11-20 02:59

每分钟刷新已保存的数据。空检查不代表采集异常;新收录可能包含历史论文。AI 导读按新增版本计数。

29,158 已收录论文24,181 AI 相关8,992 有修订112 HF 社区发现

浏览论文库

时间范围跟随排序:版本更新、本站收录或首次发表。最新收录可能包含历史论文。部分信源异常,查看状态 →

找到 33 篇论文 · 按版本更新时间排序上轮扫描起点 2026-11-21 12:17(北京时间)
版本更新 · v2arXiv:2611.22362

HealthCLBench:Evaluating Context Learning in Health Conversations through Controlled Case Clusters

Ting Yang, Zijing Hong, Chengjie Wang, Peng Wei, Chunxiao Guo, Yun Yue, Xurui Li, Jinqun Guan 等 10 位作者

原文摘要

Complex health conversations require answers grounded in individual information and currently valid evidence. We evaluate health context learning: acquiring, integrating, and applying relevant context during inference with fixed model parameters to answer health questions and respond appropriately to changes in facts, evidence, or task goals. To our knowledge, Chinese-language medical benchmarks have yet to systematically combine explicit contextual difficulties, controlled case variations, and member-specific medical criteria for open-ended answers. We introduce HealthCLBench, comprising 328 controlled clusters, 1,103 cases, and 4,931 rubric criteria across medical consultation and everyday health management. Members share backgrounds while evidence, patient states, or task goals change; each has its own medical scoring requirements. A 70-person medical team annotates and reviews the simulated cases and their rubrics. Human review of 700 full-context responses from the original evaluation yields 90.5% pooled agreement with the original automated judgments at the rubric-item level. We report full-context and query-only results for seven general-purpose models and Ant Afu. Across the eight systems, scores with full context are 11.83 to 31.76 case points and 10.87 to 38.97 geometric cluster points higher than those with the current question alone. GPT-6 Astra leads both full-context metrics at 75.03 and 60.42, followed by Claude Opus 5.5 and Ant Afu. Cross-condition differences describe score changes under the reported evaluation settings. Response comparisons reveal failures in checking historical information, expressing uncertainty when warranted, and ensuring recommendations respect required conditions. The resource supports model comparison and targeted follow-up evaluation.

cs.AI

新论文 · v1arXiv:2611.22079

SIMQ: A Single Graph-based Index for Multi-Vector Query

Yifan Lu, Jiabao Jin, Xiaoyao Zhong, Peng Cheng, Xiangyu Wang, Haotian Li, Hua Chen, Junqing Chen 等 10 位作者

原文摘要

Multi-vector retrieval models such as ColBERT achieve state-of-the-art effectiveness by representing documents as sets of token-level vectors and scoring via MaxSim, but their high computational cost limits scalability to large corpora. Existing approximate methods face fundamental bottlenecks: centroid interaction matrix approaches require computing a full matrix at every query and rely on GPU-accelerated clustering, while data-oblivious compression methods yield unstable recall that requires repeated compensation. We propose SIMQ (Single graph Index for Multi-vector Query), a framework that bridges single-vector graph indices and multi-vector retrieval through data-aware sampling-based clustering and per-token graph navigation, simultaneously bypassing all aforementioned bottlenecks. SIMQ organizes all token vectors into fine-grained clusters via data-aware sampling, builds a single graph index over cluster centroids independently of PLAID's residual representation, and approximates the MaxSim score through per-token graph search without materializing any interaction matrix or relying on GPU acceleration. We provide theoretical analysis showing that, under a manifold assumption and successful cell retrieval, data-aware partitioning has a smaller expected approximation error than a data-oblivious partition. Building on this foundation, SIMQ+ further introduces count-based scoring and a dynamic candidate set to improve robustness across varying query conditions. Experiments on five benchmarks compare recall, throughput, and construction time against several baselines, primarily on CPU; large-scale PLAID k-means centroids are built on a separate GPU machine. SIMQ has been integrated into VSAG, an Ant Group's open-source vector search library.

cs.IR · cs.DB

新论文 · v1arXiv:2611.21238

Recursive Spectral Graph Partitioning Coding for Index Assignment in Semantic Communications

Leonardo Roque, Louis-Adrien Dufrène, Quentin Lampin, Grégoire Lefebvre, Mohamad Assaad

原文摘要

Transmitting quantized semantic embeddings over a bit-flip channel requires a binary mapping that minimizes expected reconstruction distortion. The classical Binary Switching Algorithm (BSA) ($\mathcal{O}(K^3 \log K)$ for $K$ centroids) yields locally optimal assignments but ignores hierarchical bit significance, making it incompatible with Unequal Error Protection (UEP). We propose Recursive Spectral Graph Partitioning (RSGP), which recursively bisects the centroid affinity graph along the Fiedler eigenvector so that early bits separate semantically distant groups while later bits refine local structure. RSGP runs in $\mathcal{O}(K^3)$ with no iterative outer loop, and is the only method studied to show a systematic positive UEP gain: protecting early bits over later ones improves accuracy for RSGP, while being neutral or harmful for unstructured assignments. Experiments on the Banking77 dataset and the Massive Text Embedding Benchmark (MTEB) confirm that RSGP consistently matches or exceeds BSA accuracy across codebook sizes and error profiles.

cs.IT · eess.SP

新论文 · v1arXiv:2611.20758

UniGraph: A Unified Graph-based Index for Dynamic Hybrid Dense-Sparse Vector Retrieval

Keyu Chen, Xiaoyao Zhong, Jiabao Jin, Peng Cheng, Wangze Ni, Jingkuan Song, Heng Tao Shen

原文摘要

Hybrid dense-sparse vector retrieval combines semantic signals from dense representations with lexical signals from sparse representations, typically ranking items by a query-dependent weighted sum of dense and sparse vector distances. Existing systems commonly search separate dense and sparse vector indexes and rerank the union of their candidates, incurring two retrieval passes and potentially discarding items that rank highly only after the semantic and lexical signals are combined. Constructing a graph-based index directly under a hybrid distance can mitigate this premature candidate loss. However, because varying the hybrid weight alters the underlying neighborhood relationships, a single static graph built at a fixed weight cannot accommodate dynamic query-specific weights. Therefore, we propose a Unified Hybrid Graph (UniGraph), which merges neighbor sets computed at multiple hybrid weights into a single bounded-degree graph index, enabling hybrid vector retrieval with different query-specific weights without index reconstruction. We further develop an adaptive weight-guided search framework that chooses between sparse- and dense-guided entry points according to the query weight, together with a dense-first hybrid distance pruning strategy that avoids unnecessary exact sparse distance computations. Extensive experiments show that UniGraph provides a favorable Recall-QPS trade-off across different query weights. In the mixed-weight setting, UniGraph delivers 27.6%~52.3% higher QPS than the fastest baseline on each dataset at Recall@100 greater than or equal to 0.95. UniGraph has been deployed at scale in Ant Group's production environment, where hybrid retrieval with query-dependent weights is in strong demand.

cs.IR · cs.DB

版本更新 · v3arXiv:2602.19143

Incremental Learning of Sparse Attention Patterns in Transformers

Oğuz Kaan Yüksel, Rodrigo Alvarez Lucendo, Nicolas Flammarion

原文摘要

Transformers integrate information from multiple past positions through sparse attention patterns that copy from specific locations. We study how these patterns form during training in a high-order Markov-chain setting, where the next token depends on several groups of past positions with differing statistical importance. Learning proceeds in discrete stages, each marked by the emergence of a sparse attention pattern for the next most important group. The dynamics are initially competitive, with all attention heads focusing on the dominant pattern, but later become cooperative, with different heads specializing in distinct patterns. For a tractable surrogate, the dynamics reduce to those of a tensor factorization problem, enabling us to prove stagewise convergence. Empirically, the number of stages completed depends on training set size, suggesting a form of implicit complexity control through early stopping. Overall, we give a theoretical account of how structured attention patterns and head specialization emerge without an explicit curriculum.

cs.LG · math.OC · stat.ML

版本更新 · v2arXiv:2602.16140

Human-AI Collaboration in Large Language Model-Integrated Building Energy Management Systems: The Role of User Domain Knowledge and AI Literacy

Wooyoung Jung, Kahyun Jeon, Prosper Babon-Ayeng

原文摘要

This study aimed to comprehend how user domain knowledge and artificial intelligence (AI) literacy impact the effective use of human-AI interactive building energy management system (BEMS). While prior studies have investigated the potential of integrating large language models (LLMs) into BEMS or building energy modeling, very few studies have examined how user interact with such systems. We conducted a systematic role-playing experiment, where 85 human subjects interacted with an advanced generative pre-trained transformer (OpenAI GPT-4o). Participants were tasked with identifying the top five behavioral changes that could reduce home energy use with the GPT model that functioned as an LLM-integrated BEMS. Then, the collected prompt-response data and participant conclusions were analyzed using an analytical framework that hierarchically assessed and scored human-AI interactions and their home energy analysis approaches. Also, participants were classified into four groups based on their self-evaluated domain knowledge of building energy use and AI literacy, and Kruskal-Wallis H tests with post-hoc pairwise comparisons were conducted across 20 quantifiable metrics. Key takeaways include: most participants employed concise prompts (median: 16.2 words) and relied heavily on GPT's analytical capabilities; and notably, only 1 of 20 metrics, appliance identification rate, showed statistically significant group differences (p=0.037), driven by AI literacy rather than domain knowledge, suggesting an equalizing effect of LLMs across expertise levels. This study provides foundational insights into human-AI collaboration dynamics and promising development directions in the context of LLM-integrated BEMS and contributes to realizing human-centric LLM-integrated energy systems.

cs.HC · cs.AI

版本更新 · v2arXiv:2610.34682

Beyond Multiplicity Correction: Sharp Retrieval Thresholds for Attention with Correlated Distractors

Heehyun Park

原文摘要

Redundant context affects attention in two distinct ways: multiplicity adds normalization mass, while score variation within correlated groups can produce extreme individual scores. We separate these effects in a planted retrieval model with exponentially many Gaussian distractors organized into supplied clusters whose number and size both grow exponentially with dimension. The model is a two-level generalized random energy model, whose classical free energy yields exact retrieval thresholds. Gaussian interpolation further shows that increasing positive correlation at fixed token count lowers the distractor free energy, unlike adding duplicates. Our main result establishes a sharp limitation of count correction, which subtracts the log cluster size from pre-softmax scores: this correction removes multiplicity but retains within-cluster extremes. Even when the inverse temperature is chosen after observing all scores, count correction requires a signal at least $\sqrt{2α_1}$, where $α_1$ is the exponential growth rate of the number of clusters. By contrast, averaging scores before exponentiation suppresses token-level fluctuations and lowers the optimal signal threshold to $\sqrt{2ρα_1}$, where $ρ$ is the within-cluster correlation. This gap persists after adding exponentially many singleton distractors, so group size alone does not identify the target, and extends, under explicit cumulant regularity assumptions, to fixed finite hierarchies of independent-increment scores. Finite-sample, stability, and participation results complement the asymptotic analysis. A controlled retrieval experiment with frozen sentence embeddings illustrates both sides of the distinction: averaging helps when the relevant item is a singleton, but can discard useful evidence when a relevant group contains multiple matching items.

cs.LG · math.PR

版本更新 · v2arXiv:2610.26645

Reasoning-Aware Multicalibrated Reliability Predictors for Language Models

Ekaterina Fedorova, Alessandro Morosini, Chara Podimata

原文摘要

As people increasingly rely on large language models (LLMs) for information, model providers need to communicate meaningful estimates of whether individual responses are correct or, more generally, reliable. In the context of hallucinations, existing work uses multicalibration to make such reliability predictions meaningful across application-relevant groups, such as prompt topics. Topic groups alone, however, can conceal systematic errors: within the same topic and confidence level, responses with different chain-of-thought (CoT) properties may exhibit opposing errors that cancel on average. In this work, we introduce reasoning-aware multicalibration, a framework that supplements application-relevant topic groups with groups derived from observable properties of CoT traces. We highlight that, theoretically, groupings based on CoT are "internally important", i.e., groupings for which the multicalibration guarantee is not societally crucial as it is for high-risk topics, but whose inclusion in the multicalibration procedure could easily and significantly improve predictive accuracy of the final reliability predictor. Empirically, we apply our framework to hallucination prediction and unsafe-response detection across a variety of benchmarks, and find that adding CoT groups to topic-based calibration reduces Brier score in both tasks. Interestingly, we find that multicalibrating with respect to CoT groups alone (without any topic information) already improves calibration across topics, suggesting that the predictor's errors depend on how the model reasoned, rather than what it was asked about in the prompt.

cs.CY · cs.AI

首次观测 · 记录 #1ACL Anthology · ACL 2026:acl:2026.acl-long.1035ACL Anthology · ACL 2026

InsLogicBench: An Argumentation Logic Grounded Benchmark for Complex Insurance Claims Adjudication

Jin Liu, Yunpeng Liu, Keyi Wang, Jie Shi Fudan University, Xiao Xu, Wenkang Huang Ant Group, Xingzhong Xu Ant Group, Xin Liang Alibaba Group 等 9 位作者

原文摘要

Insurance claims adjudication demands not only accurate decisions but also interpretable reasoning grounded in policy clauses. However, existing benchmarks are limited to information retrieval or simple multiple-choice setups, which fail to require step-by-step inferences from facts to conclusions. To address this gap, we introduce InsLogicBench, a benchmark providing complete reasoning traces that link factual inputs, relevant policy clauses, and final verdicts. We construct the dataset using a controllable synthesis framework based on the Nested Toulmin Model. By capturing the defeasible logic of insurance policies through hierarchical truth assignment and enforcing validity via consistency verification, we ensure interpretability and logical rigor across generated examples. We evaluate eight Large Language Models (LLMs) on InsLogicBench. Results show significant difficulties in handling exception clauses and verifying missing conditions. Notably, models often produce correct final decisions but fail to provide precise justifications, highlighting a critical discrepancy between their decision accuracy and logical reasoning capabilities.

cs.CL

Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表

首次观测 · 记录 #1ACL Anthology · ACL 2026:acl:2026.acl-long.1157ACL Anthology · ACL 2026

WebSynthesis: World Model-Guided Monte Carlo Tree Search for Efficient WebAgent Trajectory Synthesis

Yifei Gao, Junhong Ye Beijing Jiaotong University, Yifan Yang, Jiaqi Wang, Yi Zhang Ant Group, Zhang Ruichen, Jitao Sang Beijing Jiaotong University

原文摘要

Recent advances in large language models (LLMs) have enabled increasingly capable web agents, yet training such agents still relies on high-quality interaction trajectories that are difficult to obtain at scale. We identify two key challenges: (1) Infrastructure Overhead, where network instability and website access restrictions limit data collection scalability; and (2) Constrained Exploration, where irreversible state transitions preclude tree-based search and thus limit trajectory diversity. To address these challenges, we introduce WebSynthesis, a framework for scalable trajectory synthesis. WebSynthesis employs an LLM-based World Model to simulate state transitions without network dependencies, and integrates Monte Carlo Tree Search to enable reversible exploration over the simulated state space. Experiments on WebArena, WebVoyager, and Mind2Web-Online demonstrate that agents trained exclusively on synthesized trajectories outperform those trained on real-world data, providing a viable alternative to costly real-world data collection.

cs.CL

Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表

首次观测 · 记录 #1ACL Anthology · ACL 2026:acl:2026.acl-long.1308ACL Anthology · ACL 2026

STAPO: Selective Trajectory-Aware Policy Optimization for LLM Agent Training

Qiuyi Qi Zhejiang University, Tian Liang Zhejiang University, Mutian Bao Zhejiang University, Jinjian Zhang Ant Group, Dongnan Liu Ant Group, Wei Zhou, Linjian Mo Ant Group, Ming Kong Zhejiang University 等 11 位作者

原文摘要

Reinforcement Learning (RL) is the dominant paradigm for training Large Language Model (LLM) agents on long-horizon tasks. However, sparse and delayed rewards often lead to trajectory neglect, in which agents lose focus on the task goal and interaction history at intermediate steps. Prior work has explored step-level supervision using Shannon-entropy–based uncertainty signals, which conflate inherent state complexity with agent confidence and therefore provide unreliable estimates of decision reliability. To address this issue, we propose normalized entropy, which measures confidence deviations relative to an agent’s average behavior under a given state, thereby strengthening the association between low-quality actions and trajectory neglect. Building on this insight, we introduce Selective Trajectory-Aware Policy Optimization (STAPO), a hierarchical group-based RL framework. STAPO leverages normalized entropy to locate outlier steps associated with trajectory neglect and optimizes them via a joint mechanism of trajectory-aware reward and trajectory-independent penalty, enhancing trajectory awareness while preserving training stability. Extensive experiments on ALFWorld, WebShop, and Search-Augmented QA demonstrate that STAPO achieves state-of-the-art performance while substantially alleviating trajectory neglect, validating its effectiveness and robustness for agentic tasks.

cs.CL · cs.AI

Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表

首次观测 · 记录 #1ACL Anthology · ACL 2026:acl:2026.acl-long.134ACL Anthology · ACL 2026

Rethinking Sample Polarity in Reinforcement Learning with Verifiable Rewards

Xinyu Tang Renmin University of China, Yuliang Zhan, Zhixun Li, Xin Zhao Renmin University of China, Zhenduo Zhang, Zujie Wen Ant Group, Zhiqiang Zhang Ant Group, Jun Zhou Ant Group

原文摘要

Large reasoning models (LRMs) are typically trained using reinforcement learning with verifiable reward (RLVR) to enhance their reasoning abilities. In this paradigm, policies are updated using both positive and negative self-generated rollouts, which correspond to distinct sample polarities. In this paper, we provide a systematic investigation into how these sample polarities affect RLVR training dynamics and behaviors. We find that positive samples sharpen existing correct reasoning patterns, while negative samples encourage exploration of new reasoning paths. We further explore how adjusting the advantage values of positive and negative samples at both the polarity level and the token level affects RLVR training. Based on these insights, we propose an Adaptive and Asymmetric token-level Advantage shaping method for Policy Optimization, namely A3PO, that more precisely allocates advantage signals to key tokens across different polarities. Experiments across five reasoning benchmarks demonstrate the effectiveness of our approach.

cs.CL

Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表

首次观测 · 记录 #1ACL Anthology · ACL 2026:acl:2026.acl-long.1359ACL Anthology · ACL 2026

EGSS: Entropy-guided Stepwise Scaling for Reliable Software Engineering

Chenhui Mao Alibaba Group, Yuanting Lei, Zhixiang Wei, Ming Liang, Zhixiang Wang Peking University, Jingxuan Xu Beijing Jiaotong University, Dajun Chen, Wei Jiang Ant Group 等 9 位作者

原文摘要

Agentic Test-Time Scaling (TTS) has delivered state-of-the-art (SOTA) performance on complex software engineering tasks such as code generation and bug fixing. However, its practical adoption remains limited due to significant computational overhead, primarily driven by two key challenges: (1) the high cost associated with deploying excessively large ensembles, and (2) the lack of a reliable mechanism for selecting the optimal candidate solution—ultimately constraining the performance gains that can be realized. To address these challenges, we propose Entropy-Guided Stepwise Scaling (EGSS), a novel TTS framework that dynamically balances efficiency and effectiveness through entropy-guided adaptive search and robust test-suite augmentation.Extensive experiments on SWE-Bench-Verified demonstrate that EGSS consistently boosts performance by 5–10% across all evaluated models. Specifically, it increases the resolved ratio of Kimi-K2-Intruct from 63.2% to 72.2%, and GLM-4.6 from 65.8% to 74.6%. Furthermore, when paired with GLM-4.6, EGSS achieves new state-of-the-art among open-source large language models. In addition to these accuracy improvements, EGSS reduces inference-time token usage by over 28% compared to existing TTS methods, achieving simultaneous gains in both effectiveness and computational efficiency.

cs.CL · cs.AI

Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表

首次观测 · 记录 #1ACL Anthology · ACL 2026:acl:2026.acl-long.1720ACL Anthology · ACL 2026

RISK: A Framework for GUI Agents in E-commerce Risk Management

Renqi Chen, Zeyin Tao Alibaba Group, Jianming Guo, Jingzhe Zhu, Yiheng Peng, Qingqing Sun, Tianyi Zhang Alipay, Shuai Chen Ant Group

原文摘要

E-commerce risk management requires aggregating diverse, deeply embedded web data through multi-step, stateful interactions, which traditional scraping methods and most existing Graphical User Interface (GUI) agents cannot handle. These agents are typically limited to single-step tasks and lack the ability to manage dynamic, interactive content critical for effective risk assessment. To address this challenge, we introduce RISK, a novel framework designed to build and deploy GUI agents for this domain. RISK integrates three components: (1) RISK-Data, a dataset of 8,492 single-step and 2,386 multi-step interaction trajectories, collected through a high-fidelity browser framework and a meticulous data curation process; (2) RISK-Bench, a benchmark with 802 single-step and 320 multi-step trajectories across three difficulty levels for standardized evaluation; and (3) RISK-R1, a R1-style reinforcement fine-tuning framework considering four aspects: (i) Output Format Constraint, (ii) Single-step and (iii) Multi-step Level Reward, and (iv) Task Level Reweight. Experiments show that RISK-R1 achieves a 6.8% improvement in offline single-step and an 8.8% improvement in offline multi-step, using only 7.2% of the parameters of the SOTA baseline. Moreover, it attains a top task success rate of 70.5% in online evaluation. RISK provides a scalable, domain-specific solution for automating complex web interactions in e-commerce risk management. The code is available at https://github.com/RenqiChen/RISK-GUI.

cs.CL

Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表

首次观测 · 记录 #1ACL Anthology · ACL 2026:acl:2026.acl-long.182ACL Anthology · ACL 2026

Can We Predict Before Executing Machine Learning Agents?

Jingsheng Zheng, Jintian Zhang, Yujie Luo, Yuren Mao, Yunjun Gao Zhejiang University, Lun Du Ant Group, Huajun Chen Zhejiang University and ZJU-Hangzhou Global Scientific and Technological Innovation Center, Ningyu Zhang Zhejiang University

原文摘要

Autonomous machine learning agents have revolutionized scientific discovery, yet they remain constrained by a Generate-Execute-Feedback paradigm. Previous approaches suffers from a severe Execution Bottleneck, as hypothesis evaluation relies strictly on expensive physical execution. To bypass these physical constraints, we internalize execution priors to substitute costly runtime checks with instantaneous predictive reasoning, drawing inspiration from World Models. In this work, we formalize the task of Data-centric Solution Preference and construct a comprehensive corpus of 18,438 pairwise comparisons. We demonstrate that LLMs exhibit significant predictive capabilities when primed with a Verified Data Analysis Report, achieving 61.5% accuracy and robust confidence calibration. Finally, we instantiate this framework in ForeAgent, an agent that employs a Predict-then-Verify loop, achieving a 6x acceleration in convergence while surpassing execution-based baselines by +6%.

cs.CL

Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表

首次观测 · 记录 #1ACL Anthology · ACL 2026:acl:2026.acl-long.1840ACL Anthology · ACL 2026

KnowRL: Exploring Knowledgeable Reinforcement Learning for Factuality

Baochang Ren, Shuofei Qiao Zhejiang University, Ningyu Zhang Zhejiang University, Da Zheng Ant Group, Huajun Chen Zhejiang University and ZJU-Hangzhou Global Scientific and Technological Innovation Center

原文摘要

Slow-thinking Large Language Models (LLMs) have demonstrated strong reasoning capabilities but often suffer from severe hallucinations due to an inability to recognize their knowledge boundaries. Existing Reinforcement Learning (RL) approaches typically rely on outcome-oriented rewards, which can inadvertently reinforce fabricated reasoning paths when the final answer is correct. To address this, we propose Knowledge-enhanced RL, KnowRL, a framework that integrates factual supervision directly into the reasoning process. By decomposing the chain of thought into atomic facts and verifying them against the corresponding ground-truth knowledge, KnowRL performs fine-grained checks to encourage models to reason faithfully. Crucially, this process-oriented supervision teaches the model to identify its knowledge boundaries, learning to say “I don’t know” instead of fabricating answers when information is missing. Experimental results demonstrate that KnowRL effectively mitigates hallucinations—reducing the Incorrect Rate on SimpleQA by 20.3% for distillation-based slow-thinking models while maintaining strong performance on complex reasoning benchmarks like GPQA and AIME 2025. Furthermore, our method shows robust transferability to out-of-distribution tasks, indicating that the model learns a generalizable verification behavior.

cs.CL

Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表

首次观测 · 记录 #1ACL Anthology · ACL 2026:acl:2026.acl-long.1867ACL Anthology · ACL 2026

UrbanGeoEval: A City-Scale Benchmark for Evaluating Large Language Models in Geospatial Reasoning

Mutian Bao Zhejiang University, Qiuyi Qi Zhejiang University, Tian Liang Zhejiang University, Jinjian Zhang Ant Group, Wei Zhou, Ming Kong Zhejiang University, Linjian Mo Ant Group, Qiang Zhu Zhejiang University

原文摘要

Current evaluations of geospatial reasoning in LLMs are frequently impeded by the entanglement of factual recall and spatial logic, which often obscures the models’ true capabilities in complex city-scale environments. To address this, we introduce UrbanGeoEval, a comprehensive benchmark featuring a dual-module framework designed to disentangle these competencies. The Knowledge Module assesses urban memory via scalable map-based queries, while the Reasoning Module isolates pure logical inference across 3,148 realistic tasks by providing necessary geospatial context. Unlike prior benchmarks that hand the model pre-computed spatial text, UrbanGeoEval provides raw geometry and forces the model to act as a spatial computing engine. Our evaluation methodology introduces a reliable hybrid pipeline that merges deterministic programmatic checks with an LLM-as-a-Judge, achieving expert-level evaluation accuracy. Extensive experiments on 18 widely used LLMs uncover critical insights: (1) models exhibit severe geographic biases and resolution gaps; (2) failures in complex multi-hop tasks often stem from brittle foundational spatial skills rather than high-level logic deficits. UrbanGeoEval provides a precise diagnostic tool for advancing urban geospatial intelligence in LLMs.

cs.CL

Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表

首次观测 · 记录 #1ACL Anthology · ACL 2026:acl:2026.acl-long.193ACL Anthology · ACL 2026

Are Emotion and Rhetoric Neurons in LLM? Neuron Recognition and Adaptive Masking for Emotion-Rhetoric Prediction Steering

Li Zheng, Xin Zhang Ant Group and Microsoft, Shuyi He, Fei Li Wuhan University, Chong Teng Wuhan University, Jiang-Ming Yang, Donghong Ji Wuhan University, Zhuang Li Royal Melbourne Institute of Technology

原文摘要

Accurate comprehension and controllable generation of emotion and rhetoric are pivotal for enhancing the reasoning capabilities of large language models (LLMs). Existing studies mostly rely on external optimizations, lacking in-depth exploration of internal representation mechanisms, thus failing to achieve fine-grained steering at the neuron level. A handful of works on neurons are confined to emotions, neglecting rhetoric neurons and their intrinsic connections. Traditional neuron masking also exhibits counterintuitive phenomena, making reliable verification of neuron functionality infeasible. To address these issues, we systematically investigate the neurons representation mechanisms and inherent associations of 6 emotion categories and 4 core rhetorical devices. We propose a neuron identification framework that integrates multi-dimensional screening, and design an adaptive masking method incorporating dynamic filtering, attenuation masking, and feedback optimization, enabling reliable functional validation of neuron functionality. Through neuron regulation, we achieve directed induction of non-target sentences and enhancement of emotion tasks via rhetoric neurons. Experiments on 5 commonly used datasets validate the effectiveness of our method, providing a novel paradigm for the fine-grained steering of emotion and rhetoric expressions in LLMs.

cs.CL

Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表

首次观测 · 记录 #1ACL Anthology · ACL 2026:acl:2026.acl-long.217ACL Anthology · ACL 2026

Instructions are all you need: Self-supervised Reinforcement Learning for Instruction Following

Qingyu Ren, Qianyu He, Powei Chang, Jie Zeng, Zeye Sun, Fei Yu Ant Group, Jiaqing Liang Fudan University, Yanghua Xiao Fudan University

原文摘要

Language models often struggle to follow multi-constraint instructions that are crucial for real-world applications. Existing reinforcement learning (RL) approaches suffer from dependency on external supervision and sparse reward signals from multi-constraint tasks. We propose a label-free self-supervised RL framework that eliminates dependency on external supervision by deriving reward signals directly from instructions and generating pseudo-labels for reward model training. Our approach introduces constraint decomposition strategies and efficient constraint-wise binary classification to address sparse reward challenges while maintaining computational efficiency. Experiments show that our approach generalizes well, achieving strong improvements across 3 in-domain and 5 out-of-domain datasets, including challenging agentic and multi-turn instruction following. We will open-source our code and data to facilitate future research.

cs.CL · cs.AI

Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表

首次观测 · 记录 #1ACL Anthology · ACL 2026:acl:2026.acl-long.354ACL Anthology · ACL 2026

Revisiting the Reliability of Language Models in Instruction-Following

Jianshuo Dong Tsinghua University, Tsinghua University, Yutong Zhang, Liu Yan, Zhenyu Zhong, Tao Wei Ant Group, Chao Zhang Tsinghua University, Han Qiu Tsinghua University

原文摘要

Advanced LLMs have achieved near-ceiling instruction-following accuracy on benchmarks such as IFEval. However, these impressive scores do not necessarily translate to reliable services in real-world use, where users often vary their phrasing, contextual framing, and task formulations. In this paper, we study nuance-oriented reliability: whether models exhibit consistent competence across cousin prompts that convey analogous user intents but with subtle nuances. To quantify this, we introduce a new metric, reliable@k, and develop an automated pipeline that generates high-quality cousin prompts via data augmentation. Building upon this, we construct IFEval++ for systematic evaluation. Across 20 proprietary and 26 open-source LLMs, we find that current models exhibit substantial insufficiency in nuance-oriented reliability—their performance can drop by up to 61.8% with nuanced prompt modifications. What’s more, we characterize it and explore three potential improvement recipes. Our findings highlight nuance-oriented reliability as a crucial yet underexplored next step toward more dependable and trustworthy LLM behavior. Our code and benchmark are accessible: https://github.com/jianshuod/IFEval-pp.

cs.CL

Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表

首次观测 · 记录 #1ACL Anthology · ACL 2026:acl:2026.acl-long.376ACL Anthology · ACL 2026

The Reasoning Trap: How Enhancing LLM Reasoning Amplifies Tool Hallucination

Chenlong Yin Pennsylvania State University, Zeyang Sha Ant Group, Shiwen Cui ant group, Changhua Meng Ant Group, Zechao Li Nanjing University of Science and Techonolgy

原文摘要

Enhancing the reasoning capabilities of Large Language Models (LLMs) is a key strategy for building agents that ”think then act”. However, recent observations, like OpenAI’s o3, suggest a paradox: stronger reasoning often coincides with increased hallucination, yet no prior work has systematically examined whether reasoning enhancement itself causes tool hallucination. We address this gap with the central question: Does strengthening reasoning increase tool hallucination? To answer this, we introduce SimpleToolHalluBench, a diagnostic benchmark measuring tool hallucination in two failure modes: (i) no tool available, and (ii) only distractor tools available. Through controlled experiments, we establish three key findings. First, we demonstrate a causal relationship: progressively enhancing reasoning through RL increases tool hallucination proportionally with task performance gains. Second, this effect transcends overfitting—training on non-tool tasks (e.g., mathematics) still amplifies subsequent tool hallucination. Third, the effect is method-agnostic, appearing when reasoning is instilled via supervised fine-tuning and when it is merely elicited at inference by switching from direct answers to step-by-step thinking. We also evaluate mitigation strategies including Prompt Engineering and Direct Preference Optimization (DPO), revealing a fundamental reliability–capability trade-off: reducing hallucination consistently degrades utility. Mechanistically, Reasoning RL disproportionately collapses tool-reliability–related representations, and hallucinations surface as amplified divergences concentrated in late-layer residual streams. These findings reveal that current reasoning enhancement methods inherently amplify tool hallucination, highlighting the need for new training objectives that jointly optimize for capability and reliability. Our implementation is provided at https://github.com/albert-y1n/Reasoning_Trap.

cs.CL

Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表

首次观测 · 记录 #1ACL Anthology · ACL 2026:acl:2026.acl-long.384ACL Anthology · ACL 2026

ReportLogic: Evaluating Logical Quality in Deep Research Reports

Jujia Zhao, Zhaoxin Huan, Zihan Wang, Xiaolu Zhang Ant Group, Jun Zhou Ant Group, Suzan Verberne Universiteit Leiden, Zhaochun Ren Leiden University

原文摘要

Users increasingly rely on Large Language Models (LLMs) for Deep Research, using them to synthesize diverse sources into structured reports that support understanding and action. In this context, the practical reliability of such reports hinges on logical quality: whether the report’s claims and arguments are explicitly supported and can be trusted as a basis for downstream use, rather than merely appearing fluent or informative. However, current evaluation frameworks largely overlook this requirement. To bridge this gap, we introduce ReportLogic, a benchmark that quantifies report-level logical quality through a reader-centric lens of auditability. Specifically, ReportLogic adopts a hierarchical taxonomy that evaluates whether readers can (1) trace an on-topic report structure with a unified analytical arc (Macro-Logic), (2) understand the progression with necessary context (Expositional-Logic), and (3) verify conclusions via explicit claim–support (Structural-Logic). Based on this taxonomy, we construct a human-annotated rubric-guided dataset and train an open-source LogicJudge for scalable evaluation. We further evaluate judge robustness via adversarial attacks, showing that off-the-shelf LLM judges are frequently influenced by superficial cues (e.g., verbosity), and reasoning modes can mask broken support relations. Overall, our results provide actionable guidance for building more robust logic evaluators and improving the logical reliability of LLM-generated reports.

cs.CL

Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表

首次观测 · 记录 #1ACL Anthology · ACL 2026:acl:2026.acl-long.504ACL Anthology · ACL 2026

Learning from the Irrecoverable: Error-Localized Policy Optimization for Tool-Integrated LLM Reasoning

Qiao Liang, Yuke Zhu Alibaba Group, Chao Ge Alibaba Group, Lei Yang Ant Group, Ying Shen Tongji University, Bo Zheng, Sheng Guo

原文摘要

Tool-integrated reasoning (TIR) enables LLM agents to solve tasks through planning, tool use, and iterative revision, but outcome-only reinforcement learning in this setting suffers from sparse, delayed rewards and weak step-level credit assignment. In long-horizon TIR trajectories, an early irrecoverable mistake can determine success or failure, making it crucial to localize the first irrecoverable step and leverage it for fine-grained credit assignment. We propose Error-Localized Policy Optimization (ELPO), which localizes the first irrecoverable step via binary-search rollout trees under a fixed rollout budget, converts the resulting tree into stable learning signals through hierarchical advantage attribution, and applies error-localized adaptive clipping to strengthen corrective updates on the critical step and its suffix. Across TIR benchmarks in math, science QA, and code execution, ELPO consistently outperforms strong Agentic RL baselines under comparable sampling budgets, with additional gains in Pass@K and Major@K scaling, rollout ranking quality, and tool-call efficiency. Our code is publicly released for reproducibility at https://anonymous.4open.science/r/ELPO-7C19.

cs.CL · cs.AI

Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表

首次观测 · 记录 #1ACL Anthology · ACL 2026:acl:2026.acl-long.524ACL Anthology · ACL 2026

LLaDA 1.5: Variance-Reduced Preference Optimization for Large Language Diffusion Models

Fengqi Zhu Renmin University of China, Rongzhen Wang Renmin University of China, Shen Nie, Xiaolu Zhang Ant Group, Chunwei Wu Ant Group, Jun Zhou Ant Group, Yankai Lin Renmin University of China, Ji-Rong Wen Renmin University of China 等 9 位作者

原文摘要

Masked diffusion language models present a promising paradigm for language modeling, yet the systematic theoretical analysis and comprehensive empirical validation of their alignment on general tasks remain relatively underexplored. In this paper, we identify the primary challenge for this problem: the high variance in Evidence Lower Bound (ELBO)-based likelihood estimates required for preference optimization. To address this issue, we propose Variance-Reduced Preference Optimization (VRPO), a framework that formally analyzes the bias and variance of the preference optimization loss and gradient based on Direct Preference Optimization, showing both are governed by a score-estimator variance. Building on this foundation, we introduce multiple unbiased variance reduction strategies, including optimal budget allocation and antithetic sampling, to improve alignment performance. We demonstrate the effectiveness of VRPO by applying it to LLaDA, a large diffusion language model. The resulting model, LLaDA 1.5, consistently outperforms its SFT-only predecessor consistently across various general benchmarks, such as mathematics (GSM8K +4.7), coding (HumanEval +3.0, MBPP +1.8), and alignment (IFEval +4.0, Arena-Hard +4.3). Furthermore, LLaDA 1.5 demonstrates a highly competitive mathematical performance compared to other strong language MDMs and ARMs. Our model is available at https://huggingface.co/GSAI-ML/LLaDA-1.5.

cs.CL

Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表

首次观测 · 记录 #1ACL Anthology · ACL 2026:acl:2026.acl-long.541ACL Anthology · ACL 2026

Fingerprinting LLMs via Prompt Injection

Yuepeng Hu, Zhengyuan Jiang Duke University, Mengyuan Li, Osama Ahmed Facebook, Zhicong Huang Ant Group, Cheng Hong Ant Group, Neil Zhenqiang Gong

原文摘要

Large language models (LLMs) are often modified after release through post-processing such as post-training or quantization, which makes it challenging to determine whether one model is derived from another. Existing provenance detection methods have two main limitations: (1) they embed signals into the base model before release, which is infeasible for already published models, or (2) they compare outputs across models using hand-crafted or random prompts, which are not robust to post-processing. In this work, we propose LLMPrint, a novel detection framework that constructs fingerprints by exploiting LLMs’ inherent vulnerability to prompt injection. Our key insight is that by optimizing fingerprint prompts to enforce consistent token preferences, we can obtain fingerprints that are both unique to the base model and robust to post-processing. We further develop a unified verification procedure that applies to both gray-box and black-box settings, with statistical guarantees. We evaluate LLMPrint on five base models and around 700 post-trained or quantized variants. Our results show that LLMPrint achieves high true positive rates while keeping false positive rates near zero. The code is publicly available at https://github.com/hifi-hyp/ACL-LLMPrint.

cs.CL

Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表

首次观测 · 记录 #1ACL Anthology · ACL 2026:acl:2026.acl-long.821ACL Anthology · ACL 2026

Improving Autoformalization Using Direct Dependency Retrieval

Shaoqi Wang, Lu Yu Ant Group, Siwei Lou, Feng Yan, Chunjie Yang Zhejiang University, Qing Cui, Jun Zhou Ant Group

原文摘要

Statement autoformalization, a crucial first step in formal verification, aims to transform informal descriptions of math problems into machine-verifiable formal representations but remains a significant challenge. The core difficulty lies in the fact that existing language models hallucinate formal dependencies, including missing or incorrect definitions, lemmas, and theorems. Current dependency retrieval approaches exhibit poor precision and recall, and lack the scalability to leverage ever-growing public datasets. To bridge this gap, we propose a novel retrieval-augmented framework based on Direct Dependency Retrieval (DDR). DDR directly generates candidate formal dependencies from natural-language mathematical descriptions and verifies their existence in the formal library via an efficient Suffix Array Check (SAC). Built on a SAC-constructed dependency retrieval dataset of over 500,000 samples, a high-precision DDR model is fine-tuned and shown to significantly outperform state-of-the-art methods in both retrieval precision and recall, leading to superior advantage in the autoformalization tasks. SAC also contributes in assessing formalization difficulty and enabling explicit quantification of the hallucination in In-Context Learning (ICL).

cs.CL

Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表

首次观测 · 记录 #1ACL Anthology · ACL 2026:acl:2026.acl-long.988ACL Anthology · ACL 2026

WebClipper: Efficient Evolution of Web Agents with Graph-based Trajectory Pruning

Junjie Wang Ant Group, Zequn Xie, Dan Yang, Jie Feng, Yue Shen antgroup, Duolin Sun, Meixiu Long, Yihan Jiao 等 12 位作者

原文摘要

Deep Research systems based on web agents have shown strong potential in solving complex information-seeking tasks, yet their search efficiency remains underexplored. We observe that many state-of-the-art open-source web agents rely on long tool-call trajectories with cyclic reasoning loops and exploration of unproductive branches. To address this, we propose WebClipper, a framework that compresses web agent trajectories via graph-based pruning. Concretely, we model the agent’s search process as a state graph and cast trajectory optimization as a minimum-necessary Directed Acyclic Graph (DAG) mining problem, yielding pruned trajectories that preserve essential reasoning while eliminating redundant steps. Continued training on these refined trajectories enables the agent to evolve toward more efficient search patterns and reduces tool-call rounds by about 20% while improving accuracy. Furthermore, we introduce a new metric called F-AE Score to measure the model’s overall performance in balancing accuracy and efficiency. Experiments demonstrate that WebClipper compresses tool-call rounds under excellent performance, providing practical insight into balancing effectiveness and efficiency in web agent design.

cs.CL

Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表

首次观测 · 记录 #1ACL Anthology · ACL 2026:acl:2026.acl-long.999ACL Anthology · ACL 2026

Perplexity-Aware Data Scaling Law: Perplexity Landscapes Predict Performance for Continual Pre-training

Lei Liu Alibaba Group, Hao Zhu Ant Group, Xiaoyan Yang, Yue Shen antgroup, Zhixuan Chu Zhejiang University, Jian Wang, Jinjie Gu, Kui Ren Shanghai Institute for Advanced Study, Zhejiang University, The State Key Laboratory of Blockchain and Data Security, Zhejiang University and Zhejiang University

原文摘要

Continual Pre-training (CPT) serves as a fundamental approach for adapting foundation models to domain-specific applications. Scaling laws for pre-training define a power-law relationship between dataset size and the test loss of an LLM. However, the marginal gains from simply increasing data for CPT diminish rapidly, yielding suboptimal data utilization and inefficient training. To address this challenge, we propose a novel perplexity-aware data scaling law to establish a predictive relationship between the perplexity landscape of domain-specific data and the test loss. Our approach leverages the pre-trained model’s own perplexity on domain data as a proxy for estimating the knowledge gap, effectively quantifying the informational perplexity landscape of candidate training samples. By fitting this scaling law across diverse perplexity regimes, we enable adaptive selection of high-utility data subsets, prioritizing content that maximizes knowledge absorption while minimizing redundancy and noise. Extensive experiments on both medical and general-domain benchmarks demonstrate that our method consistently identifies near-optimal training subsets, achieving superior performance with significantly reduced data consumption.

cs.CL

Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表

首次观测 · 记录 #1ACL Anthology · ACL 2026:acl:2026.acl-short.12ACL Anthology · ACL 2026

StructMem: Structured Memory for Long-Horizon Behavior in LLMs

Buqiang Xu, Yijun Chen, Jizhan Fang, Ruobin Zhong Zhejiang University, Yunzhi Yao Zhejiang University, Yuqi Zhu Zhejiang University, Lun Du Ant Group, Shumin Deng Zhejiang University

原文摘要

Long-term conversational agents need memory systems that capture relationships between events, not merely isolated facts, to support temporal reasoning and multi-hop question answering. Current approaches face a fundamental trade-off: flat memory is efficient but fails to model relational structure, while graph-based memory enables structured reasoning at the cost of expensive and fragile construction. To address these issues, we propose StructMem, a structure-enriched hierarchical memory framework that preserves event-level bindings and induces cross-event connections. By temporally anchoring dual perspectives and performing periodic semantic consolidation, StructMem improves temporal reasoning and multi-hop performance on LoCoMo, while substantially reducing token usage, API calls, and runtime compared to prior memory systems.

cs.CL

Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers) · 2026 · 正式发表

首次观测 · 记录 #1ACL Anthology · ACL 2026:acl:2026.acl-short.70ACL Anthology · ACL 2026

What Makes AI Research Replicable? Executable Knowledge Graphs as Scientific Knowledge Representations

Yujie Luo, Zhuoyun Yu, Xuehai Wang, Yuqi Zhu Zhejiang University, Ningyu Zhang Zhejiang University, Lanning Wei AntGroup, Lun Du Ant Group, Da Zheng Ant Group 等 9 位作者

原文摘要

Replicating AI research is a crucial yet challenging task for large language model (LLM) agents. Existing approaches often struggle to generate executable code, primarily due to insufficient background knowledge and the limitations of retrieval-augmented generation (RAG) methods, which fail to capture latent technical details hidden in referenced papers. Furthermore, previous approaches tend to overlook valuable implementation-level code signals and lack structured knowledge representations that support multi-granular retrieval and reuse. To overcome these challenges, we propose Executable Knowledge Graphs (xKG), a pluggable, paper-centric knowledge base that automatically integrates code snippets and technical insights extracted from scientific literature. When integrated into three agent frameworks with two different LLMs, xKG shows substantial performance gains (10.9% with o3-mini) on PaperBench, demonstrating its effectiveness as a general and extensible solution for automated AI research replication.

cs.CL

Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers) · 2026 · 正式发表