arXiv cs.AI· Chenxiao Fan, Chongming Gao, Gangyi Zhang, Yaxin Gong, Leyang Shen, Jiakai Wang, Dong Wang, Yang Liu, Haoyan Liu, Xinting Hu, Fuli Feng, Xiangnan He·· 14 小时前AI 评分31
MASCAL：为多智能体协作学习校准置信度
Learning Calibrated Confidence for Multi-Agent Collaboration
AI 导读
研究者提出 MASCAL，让多智能体系统中的每个智能体通过强化学习联合优化任务表现与置信度校准，为自身答案输出正确概率，并在交接时传给下游智能体作为量化判断依据。该方法还引入置信度感知加权，提升下游智能体对上游概率信号的利用。在覆盖 Web 检索与推理任务的三种协作架构上，MASCAL 同时提升了双方智能体的任务准确率与校准质量，并可泛化到更深的协作链和不同模型规模。
正文 · 原文
Authors:Chenxiao Fan, Chongming Gao, Gangyi Zhang, Yaxin Gong, Leyang Shen, Jiakai Wang, Dong Wang, Yang Liu, Haoyan Liu, Xinting Hu, Fuli Feng, Xiangnan He
Abstract:Complex information seeking on the Web often requires collecting evidence from multiple documents and reasoning across them, as in multi-hop question answering. Multi-agent systems address such tasks by dividing work among specialized agents, where upstream agents produce intermediate answers and supporting evidence that downstream agents build upon. However, upstream answers vary in correctness, and most systems provide no quantitative signal about it, leaving the downstream agent to judge whether to trust or verify what it receives based solely on the textual evidence.
In this paper, we propose MASCAL. Each agent learns to produce a correctness probability for its answer by jointly optimizing task performance and confidence calibration in reinforcement learning; the upstream agent passes this probability at handoff, giving the downstream agent a quantitative basis for its decisions, and the final answer also carries a calibrated probability, quantifying the reliability of the system output. To guide the downstream agent in using this signal during training, we introduce confidence-aware weighting, which increases the learning weight at positions responsive to the upstream probability, so that the agent learns more from decisions where it actively uses the signal. Across three collaboration architectures spanning Web retrieval and reasoning tasks, MASCAL improves task accuracy and calibration quality for both agents, and generalizes to deeper collaboration chains and different model scales. Intervention analysis confirms that upstream confidence affects downstream decision behavior, and the learned final confidence supports selective answering and targeted rechecking.
| Subjects: | Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA) |
| Cite as: | arXiv:2610.39927 [cs.LG] |
| (or arXiv:2610.39927v1 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2610.39927 arXiv-issued DOI via DataCite |
Submission history
From: Chenxiao Fan [view email]
[v1]
Thu, 29 Oct 2026 12:30:00 UTC (322 KB)
来源：arXiv cs.AI · arxiv.org