PRIME: Towards Verifiable Evidence-Grounded Reasoning in LLM Agents

查看原文 / View Original 编辑 / Edit

Computer Science > Artificial Intelligence arXiv:2610.22426 (cs) [Submitted on 15 Oct 2026] Title:PRIME: Towards Verifiable Evidence-Grounded Reasoning in LLM Agents Authors:Hamed Khosravi, Xiaoming Huo View a PDF of the paper titled PRIME: Towards Verifiable Evidence-Grounded Reasoning in LLM Agents, by Hamed Khosravi and 1 other authors View PDF HTML (experimental)

计算机科学 > 人工智能 arXiv:2610.22426 (cs) [提交于 2026 年 10 月 15 日] 标题：PRIME：迈向大语言模型代理中可验证的证据基础推理作者：Hamed Khosravi, Xiaoming Huo 查看题为《PRIME：迈向大语言模型代理中可验证的证据基础推理》的论文 PDF，作者：Hamed Khosravi 及其他 1 位作者查看 PDF HTML（实验性）

Abstract:LLM agents are used to check records against an organization’s requirements, for example whether a vulnerability report is ready for investigation. Requirements often allow several ways to be met, such as a severity rating with a curator-reviewed advisory or a published advisory link. Agents fail by accepting a report after a required check has failed, or by settling for a weaker conclusion without trying another way that could still succeed.

摘要：大语言模型（LLM）代理被用于根据组织的要求检查记录，例如漏洞报告是否已准备好进行调查。要求通常允许通过多种方式满足，例如带有策展人审查公告的严重性评级或已发布的公告链接。代理的失败表现为在必要检查失败后接受报告，或者在不尝试其他可能成功的方法的情况下满足于较弱的结论。

We introduce PRIME (Precommitted Rules for Inference from Minimal Evidence). Before checking, it fixes evidence routes specifying how each conclusion may be justified. A separate program records each check result and refuses any conclusion the results do not justify. It also points the agent to checks that could justify a stronger conclusion, or runs them itself.

我们引入了 PRIME（基于最小证据推理的预承诺规则）。在检查之前，它会固定证据路径，明确每个结论如何被证明。一个独立的程序会记录每次检查的结果，并拒绝任何结果无法证明的结论。它还会引导代理进行可能证明更强结论的检查，或者自行运行这些检查。

PRIME reports only conclusions authorized by its rules and ledger. With correct requirements, exact checks at known locations, and sufficient budget at stopping, completion reaches the strongest supported conclusion. On 480 public-record tasks, an agent shown the requirements reported unjustified conclusions on 26.7% of tasks, while PRIME reported none. Its directions and added checks raised the average score for justified conclusions from 30.1, with refusal alone, to the maximum of 59.2 out of 100.

PRIME 仅报告其规则和账本授权的结论。在要求正确、已知位置检查准确且停止时预算充足的情况下，完成过程会得出最强有力的支持结论。在 480 项公共记录任务中，一个被展示了要求的代理在 26.7% 的任务中报告了未经证实的结论，而 PRIME 未报告任何此类结论。其指导和增加的检查将合理结论的平均分从仅使用拒绝时的 30.1 分提高到了满分 100 分中的 59.2 分。