Long-running LLM agents compact their context, replacing most history with a summary. We ask whether handoff summaries state permissions or requirements the record does not support, whether such sentences change what the next instance (the successor) does, and where they come from. Synthetic tool-use tasks carry a ground-truth permission ledger; same-snapshot replay reruns the successor with those whole sentences replaced by neutral filler. Two AI readers, reading only rule-flagged held-out summaries, confirmed unsupported restrictions in 1.2% and 2.4% of all readable summaries from two open-weight models, and unsupported authorizations in 0.1% and 1.3%. For restrictions, a pre-specified replay test was not significant (p = 0.58); a follow-up, planned after it, found a fragile effect (p = 0.027). Neutralizing false approvals cut unauthorized actions from 24 to 7 of 120 held-out runs (post hoc; later set: 17 to 11 of 48). Most development restrictions followed the agent's earlier reasoning that an action needs approval (post hoc). In pre-specified tests, replacing the system prompt's "Some actions need approval." with one saying where requirements come from cut unsupported restrictions from 5.0% to 1.8% of development summaries and, on held-out templates, from 1.8% to 0.3% (both p < 10^-9), while requests for approvals the ledger requires barely changed. Code and data: https://github.com/Chuw1/handoff-audit
He Geng, Yangmin Huang IFLYTEK CO.LTD., Lixian Lai Xunfei Healthcare, Qianyun Du IFLYTEK CO.LTD., Hui Chu, Zhiyang He Xunfei Healthcare Technology Co., Ltd., Jiaxue Hu IFLYTEK CO.LTD., Xiaodong Tao Xunfei Healthcare Co. Ltd
Aligning Large Language Models (LLMs) with high-stakes medical standards remains a significant challenge, primarily due to the dissonance between coarse-grained preference signals and the complex, multi-dimensional nature of clinical protocols. To bridge this gap, we introduce \textit{ProMedical}, a unified alignment framework grounded in fine-grained clinical criteria. We first construct \textit{ProMedical-Preference-50k}, a dataset generated via a human-in-the-loop pipeline that augments medical instructions with rigorous, physician-derived rubrics. Leveraging this corpus, we propose the Explicit Criteria Injection paradigm to train a multi-dimensional reward model. Unlike traditional scalar reward models, our approach explicitly disentangles safety constraints from general proficiency, enabling precise guidance during reinforcement learning. To rigorously validate this framework, we establish \textit{ProMedical-Bench}, a held-out evaluation suite anchored by double-blind expert adjudication. Empirical evaluations demonstrate that optimizing the Qwen3-8B base model via \textit{ProMedical-RM}-guided GRPO yields substantial gains, improving overall accuracy by 22.3% and safety compliance by 21.7%, effectively rivaling proprietary frontier models. Furthermore, the aligned policy generalizes robustly to external benchmarks, demonstrating performance comparable to state-of-the-art models on UltraMedical. We publicly release our datasets, reward models, and benchmarks to facilitate reproducible research in safety-aware medical alignment.
尚无 AI 解读，展示原文摘要
cs.CL
Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表
Zhen Wan, Chao-Han Huck Yang NVIDIA Research, Jinchuan Tian, Hanrong Ye NVIDIA, Ankita Pasad NVIDIA, Szu-Wei Fu NVIDIA and Microsoft, Arushi Goel NVIDIA Research, Ryo Hachiuma NVIDIA 等 18 位作者
We introduce a voice-agentic framework that learns one critical omni-understanding skill: knowing when to trust itself versus when to consult external audio perception. Our work is motivated by a crucial yet counterintuitive finding: naively fine-tuning an omni-model on both speech recognition and external sound understanding tasks often degrades performance, as the model can be easily misled by noisy hypotheses. To address this, our framework, Speech-Hands, recasts the problem as an explicit self-reflection decision. This learnable reflection primitive proves effective in preventing the model from being derailed by flawed external candidates. We show that this agentic action mechanism generalizes naturally from speech recognition to complex, multiple-choice audio reasoning. Across the OpenASR leaderboard, which includes seven domain-diverse speech datasets, Speech-Hands consistently outperforms strong baselines by 12.1% WER on the OpenASR benchmark. The model also achieves 77.37% accuracy and high F1 on audio QA decisions, showing robust generalization and reliability across diverse audio question answering datasets. By unifying perception and decision-making, our work offers a practical path toward more reliable and resilient audio intelligence.
尚无 AI 解读，展示原文摘要
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
Prompt sensitivity, which refers to how strongly the output of a large language model (LLM) depends on the exact wording of its input prompt, raises concerns among users about the LLM’s stability and reliability. In this work, we consider LLMs as multivariate functions and perform a first-order Taylor expansion, thereby analyzing the relationship between meaning-preserving prompts, their gradients, and the log probabilities of the model’s next token. We derive an upper bound on the difference between log probabilities using the Cauchy-Schwarz inequality. We show that LLMs do not internally cluster similar inputs like smaller neural networks do, but instead disperse them. This dispersing behavior leads to an excessively high upper bound on the difference of log probabilities between two meaning-preserving prompts, making it difficult to effectively reduce to 0. In our analysis, we also show which types of meaning-preserving prompt variants are more likely to introduce prompt sensitivity risks in LLMs. In addition, we demonstrate that the upper bound is strongly correlated with an existing prompt sensitivity metric, PromptSensiScore. Moreover, by analyzing the logit variance, we find that prompt templates typically exert a greater influence on logits than the questions themselves. Overall, our results provide a general interpretation for why current LLMs can be highly sensitive to prompts with the same meaning, offering crucial evidence for understanding the prompt sensitivity of LLMs.
尚无 AI 解读，展示原文摘要
cs.CL
Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表
Evaluating meeting effectiveness is crucial for improving organizational productivity. Current approaches rely on post-hoc surveys that yield a single coarse-grained score for an entire meeting. The reliance on manual assessment is inherently limited in scalability, cost, and reproducibility. Moreover, a single score fails to capture the dynamic nature of collaborative discussions. We propose a new paradigm for evaluating meeting effectiveness centered on novel criteria and temporal fine-grained approach. We define effectiveness as the rate of objective achievement over time and assess it for individual topical segments within a meeting. To support this task, we introduce the AMI Meeting Effectiveness (AMI-ME) dataset, a new meta-evaluation dataset containing 2,459 human-annotated segments from 130 AMI Corpus meetings. We also develop an automatic effectiveness evaluation framework that uses a Large Language Model (LLM) as a judge to score each segment’s effectiveness relative to the overall meeting objectives. Through substantial experiments, we establish a comprehensive benchmark for this new task and evaluate the framework’s generalizability across distinct meeting types, ranging from business scenarios to unstructured discussions. Furthermore, we benchmark end-to-end performance starting from raw speech to measure the capabilities of a complete system. Our results validate the framework’s effectiveness and provide strong baselines to facilitate future research in meeting analysis and multi-party dialogue. Our dataset and code will be publicly available.
尚无 AI 解读，展示原文摘要
cs.CL
Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表
Qian Zhang, Xinye Li, Xiaokai Wu, Junhao Xu, Zhanyue Qin Harbin Institute of Technology (Shenzhen) and Tencent, Qingbin Liu, Junxian Cai Tencent, Xi Chen 等 13 位作者
Model Editing, also known as knowledge editing, is receiving increasing attention in the field of Large Language Models (LLMs). However, existing model editing approaches predominantly focus on knowledge-level or static visual domains, overlooking dynamic semantics. This paper exploratively applies six representative model editing methods (FT, IKE, MEND, SERAC, MEMIT and AlphaEdit) to Video Large Language Models (Vid-LLMs) and introduces the first benchmark specifically designed for Vid-LLMs editing—VMEB (Vid-LLMs Model Editing Benchmark)—systematically extending model editing research from static modalities to dynamic video scenarios. We position this work as a forward-looking benchmark and a foundational diagnostic study: in the video paradigm, our evaluation dimensions encompass traditional metrics including Reliability, Locality, and Generality, while also introducing a video-specific metric: Robustness. Based on experimental results, we analyze the strengths and limitations of existing model editing approaches, and identify new challenges and research directions for the future development of the model editing field within the context of multimodal and video paradigms. Our benchmark is available at https://github.com/Sakabamrisa/VMEB.
尚无 AI 解读，展示原文摘要
cs.CL · cs.CV
Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表