arXiv 论文：被扭曲的工具格式化输出如何误导 Agent 及其验证器
Who Verifies the Tools? How Warped Formatters Mislead Agents and Their Verifiers
AI 导读
arXiv 论文（arXiv:2610.32110，NeurIPS 2026 workshop poster）研究工具输出格式化文本如何误导 Agent 与验证器，起因是 Pilsudski Institute 生产 RAG 系统中各工具计算正确但输出省略搜索限制、结果上限或标识符命名空间。
正文 · 原文
Abstract:Agents and verifiers often rely on the text a tool prints, and that text can misrepresent what the tool actually did, even when the tool's computation is correct. We saw this happen in a production RAG research system used by the Pilsudski Institute in New York, where tool output formatters misled an agent into giving wrong or incomplete information to users. Each tool did its job correctly, but its output omitted search limits, result caps, or identifier namespaces. We also recreated an unlabeled estimate caught in code review before release. We studied whether these formatters misled agents and verifiers and how much the repaired wording helps. We fed tool results to agents using two formatter versions: one with the key omissions and one with all missing details clearly stated. That difference alone changed what the agents claimed and what the verifiers accepted. Unsupported agent claims dropped substantially for three of the four failure types. In one test, verifiers went from accepting incorrect answers 90 out of 90 times to rejecting them 90 out of 90 times. A control showed the verifiers could catch the errors once they were given the missing facts and asked to judge correctness. We include the audit prompt we used with Codex and Claude Code in an exploratory demonstration, written with prior knowledge of our own system's defects. When pointed at an old version of our own system, it rediscovered the defects that were still live and surfaced twelve more we had not catalogued. When pointed at OpenClaw and Hermes, two of the best-known open-source agentic systems, it found warped formatters in both. We opened PRs to fix both, and the OpenClaw fix has already been merged. Practitioners can run the same prompt on their own systems today.
| Comments: | Accepted as a poster at the Who Verifies the Agents? workshop at NeurIPS 2026 (non-archival). 12 pages, 3 figures. Code and data: this https URL |
| Subjects: | Artificial Intelligence (cs.AI); Software Engineering (cs.SE) |
| Cite as: | arXiv:2610.32110 [cs.AI] |
| (or arXiv:2610.32110v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2610.32110 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: William Jeffries [view email]
[v1]
Fri, 23 Oct 2026 03:22:12 UTC (41 KB)
来源：arXiv cs.AI · arxiv.org