arXiv cs.AI· Iftach Shoham, Yuval David, Offer Akrabi, Erez Bilgory, Ido Levy, Sami Marreed, Sergey Zeltyn, Segev Shlomov·· 2 小时前AI 评分47
Think Before You Act：面向 LLM 智能体的决策门框架 JevAct
Think Before You Act: Decision Gate for LLM Agents
AI 导读
一篇 arXiv 论文提出 JevAct，一个运行时验证框架，可在 LLM 智能体动作执行前拦截并进行反馈驱动修正。该框架将领域策略编译为可执行状态检查和结构化问题，由仅 prefill 的决策模型给出基于概率的判断，不依赖 LLM 生成，也不修改底层智能体的参数或工具。
正文 · 原文
Abstract:LLM agents increasingly perform consequential actions through external tools, yet their proposed actions are often executed without independent verification of policy compliance or task completion. We introduce JevAct, a runtime verification framework that intercepts selected agent actions before execution and enables feedback-driven correction. JevAct compiles domain policies into executable state checks and structured questions evaluated by a prefill-only decision model, which produces probability-based judgments without relying on LLM generation. By combining deterministic verification of tool-state facts with probabilistic assessment of conversational context, the framework identifies potentially invalid actions and premature task termination, providing targeted feedback for bounded revision. JevAct operates independently of the underlying agent model and requires no modifications to its parameters or tools. We evaluate JevAct across eleven LLM agents on $\tau^2$-bench Airline and Telecom tasks and BFCL. Results show improvements in agent reliability across multiple models and benchmarks, with particularly consistent gains on Telecom and BFCL, while performance on Airline is more mixed. Overall, these findings suggest that combining structured decision models with executable verification can improve the reliability of tool-using LLM agents.
| Subjects: | Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2610.38620 [cs.AI] |
| (or arXiv:2610.38620v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2610.38620 arXiv-issued DOI via DataCite |
Submission history
From: Segev Shlomov [view email]
[v1]
Wed, 28 Oct 2026 10:40:52 UTC (45 KB)
来源：arXiv cs.AI · arxiv.org