CLEAT (CLarification through valuE-guided Agent co-Training) is a framework that jointly trains a clarification-value controller and a task policy to improve interactive intent alignment in large language model agents. It introduces a residual query value metric, estimated via paired rollouts, to assess the utility of asking clarification questions and uses this to guide both controller predictions and policy decisions. Experiments with Qwen3-4B show CLEAT outperforming baselines across multiple benchmarks, with significant gains in user‑centric tasks and demonstrating that co‑training benefits the task policy even when the controller is omitted at inference.
By Jiatan Huang, Zheyuan Zhang, Ziming Li, Tianyi Ma, Mingchen Li, Yanfang Ye, Chuxu Zhang
The paper investigates strategic clarification in dialogue, focusing on when large language models should ask clarification questions versus providing immediate answers to resolve ambiguous references. Using a controlled test environment derived from the SIMMC 2.1 dataset, the authors evaluate baseline LLM behavior and post‑training methods, noting a bias toward premature answering. They introduce BACE, a lightweight belief‑aware critic that tracks epistemic uncertainty, and demonstrate that it more consistently selects the optimal number of clarifying questions compared to methods like GRPO and ArCHer.
By Ibrahim Khebour, Nikhil Krishnaswamy
The paper introduces VagueBench, a benchmark that degrades seed queries into vague variants to evaluate large language models’ ability to ask follow‑up questions when user queries are incomplete. It proposes DialogGRPO, an offline reinforcement learning approach that models multi‑turn elicitation as a dialogue‑state directed acyclic graph, using dense turn‑level rewards to guide question selection. DialogGRPO achieves over 88% success in eliciting missing information on vague queries, improves answer coverage and recovery efficiency, and outperforms prompting and test‑time scaling baselines while generalizing zero‑shot to other dialogue datasets.
By Rongyu Zhang, Mutian Bao, Jinjian Zhang, Yuetian Du, He Xu, Wei Zhou, Xiaojun Zhu, Ming Kong, Jie Liu, Linjian Mo, Qiang Zhu
arXiv:2608. 12764v1 Announce Type: cross Abstract: Deep search agents operate over trajectories spanning dozens of steps, yet standard reinforcement learning provides only a single outcome reward per trajectory, which is far too sparse for effective credit assignment.
By Haoze Wu, Chuqiao Kuang, Tianyi Zhuang, Xiaoguang Li
arXiv:2606. 03135v1 Announce Type: new Abstract: Large Language Model (LLM) agents often operate under underspecified user instructions, where latent uncertainty over user intent leads to erroneous tool actions.
By Mengyi Deng, Zhiwei Li, Xin Li, Tingyu Zhu, Ying Zhao, Zhijiang Guo, Wei Wang
The paper introduces VAKE, a two‑stage reinforcement‑learning framework that activates latent factual knowledge in large language models. In the Priming stage, the model explicitly inserts bridging triples into an insufficient subgraph, guided by rewards from a frozen model’s answers. The Reasoning stage then trains the model to answer from the original input, demonstrating that the elicitation capability transfers to implicit reasoning and consistently outperforms baselines across multiple benchmarks and model sizes.
By Zuocheng Ying, Yang Yang, Yumou Wu, Chuanbo Zhu, Jiarui Wang, Ziqi Wu, Jingming Cai, Junqing Yu, Zikai Song
arXiv:2607.28077v2 Announce Type: replace
Abstract: Reinforcement learning with verifiable rewards (RLVR) improves the reasoning capabilities of large language models, but prompt groups with identica...
By Shuang Liang, Haoyang Zhou, Yifan Gong, Guowei Wang, Xiting Wang
The paper investigates how reinforcement learning can cause large language model agents to adopt shortcut policies for tool use, relying on superficial prompt cues rather than actual task needs. By creating synthetic environments that mix factual QA and math reasoning, the authors show that agents often invoke tools when cues are present, even when those tools are unnecessary, with spurious invocation rates rising up to 39%. They find that shortcut learning occurs mainly when agents have already mastered the target tool and that semantic alignment between cues and tools amplifies the effect. To counter this, they propose a dense, decision-level reward where an LLM judge assesses tool necessity, which reduces cue-driven tool use while maintaining performance.
By Yiwei Yang, Haoxiang Zhang, Bingbing Wen, Yao Lu, Yuchen Wu, Lei Zhang, Julian McAuley, Pan Lu, Bill Howe
COFFEQA introduces a counterfactual-guided few‑shot distillation method for open‑book question answering, leveraging minimally modified contexts that prompt the teacher model to produce alternative answers. By pairing factual samples with validated counterfactuals—filtered using a new Context Following Difficulty metric—COFFEQA constrains the student’s decision boundary to better match the teacher’s. Experiments on multiple QA benchmarks show that COFFEQA consistently outperforms factual‑only distillation, achieving significant gains in EM and F1 scores with a limited few‑shot budget.
By Kavinda Kehelella, Sanghamitra Dutta
arXiv:2605.25831v2 Announce Type: replace-cross
Abstract: Large language models (LLMs) define a distribution over text, which can be viewed as a probabilistic representation of uncertainty: sampling...
By Joris Baan, Wilker Aziz, Barbara Plank, Raquel Fern\'andez
MEA is a multi‑agent framework that automates the selection and configuration of explanation tools for machine learning models across tabular, text, and vision data. A Proposer agent chooses appropriate tools based on the question and modality, while an Actor agent is trained end‑to‑end for faithfulness, producing natural‑language explanations. Experiments show that MEA outperforms existing post‑hoc explainers and closed‑source baselines, achieving significant faithfulness gains on six datasets.
By Yuyang Cheng, Raghav Kaushik Ravi, Srivarshinee Sridhar, Sriparna Saha, Akash Ghosh, Chirag Agarwal
Reinforcement learning with verifiable rewards (RLVR) improves the reasoning capabilities of large language models, but prompt groups with identical rollout rewards consume generation budget without effective learning signals. Pre-rollout prompt selection can reduce this waste by screening prompts before rollout generation.