跳到正文
研踪PaperTrail
← 研踪 PaperTrail

arXiv:2609.32495 · v2

Hearsay: Who Can Audit the Record a Deployed Agent Harness Writes?

首次提交 2026-09-26 19:37
版本更新 2026-11-18 20:50
雷达观测 2026-11-19 15:04
arXiv ↗PDF ↗

摘要

The record of an agent's run is written by the harness that drives the agent. When the run is later disputed, the auditor was not there and has only this record. We ask whether someone who was not there can check the record, and what decides the answer. It depends on who wrote what a check can rely on: if the harness wrote all of it, the record is only the harness's own word, and we call it hearsay. We build Hearsay, a benchmark that examines the record and not the task: five harnesses run fourteen tasks, model examiners from three vendors and a human panel read the records and answer questions about the runs, and every passage they cite is labelled mechanically by who wrote it. We also audit what is deployed: in all sixteen frameworks, the harness is the only writer. With one writer, readers can name the fault from the record, in up to nine runs in ten, but they cannot check the record. The model examiners claim corroboration for 18 to 55 points more of their evidence than the labels support, a gap we then predicted for runs driven by another vendor's model and found again. And the best of them notices only half of the alterations we make to a record, while the two that name the fault most often cannot tell what a harness left out. With a second writer, the record can be checked. Our sandbox keeps a ledger of what passes between harness and model, and reading the ledger against the record in both directions shows what the record leaves out and what it adds. We test this as an auditor would use it, on one unseen record at a time. On runs driven by a third vendor's model, with every rule committed before the records existed, it reports 64 of 66 alterations and flags none of 84 clean records. On held-out records in five formats it reports 109 of 115. What makes a record evidence is a second writer, not a better reader.

作者备注:50 pages (10-page main text), 8 figures. Under review. v2: re-organized around one question; the second writer is now evaluated one unseen record at a time, on held-out records and under rules committed before the records existed; title changed. Code and data: https://anonymous.4open.science/r/hearsay-4E7C

AI 论文解读

基于摘要 · 非全文审阅

此版本尚未生成 AI 解读。解读由后台任务生成并按论文版本保存,登录后共享已有结果。

版本差异

本站首次收录为 v2,暂无更早版本可对照。

观测到的版本

相邻记录可能跳过中间版本;首次发现 v2 及以上也会标为版本更新。

v2 · 2026-11-18 20:50 · 最新观测版本

Hearsay: Who Can Audit the Record a Deployed Agent Harness Writes?

The record of an agent's run is written by the harness that drives the agent. When the run is later disputed, the auditor was not there and has only this record. We ask whether someone who was not there can check the record, and what decides the answer. It depends on who wrote what a check can rely on: if the harness wrote all of it, the record is only the harness's own word, and we call it hearsay. We build Hearsay, a benchmark that examines the record and not the task: five harnesses run fourteen tasks, model examiners from three vendors and a human panel read the records and answer questions about the runs, and every passage they cite is labelled mechanically by who wrote it. We also audit what is deployed: in all sixteen frameworks, the harness is the only writer. With one writer, readers can name the fault from the record, in up to nine runs in ten, but they cannot check the record. The model examiners claim corroboration for 18 to 55 points more of their evidence than the labels support, a gap we then predicted for runs driven by another vendor's model and found again. And the best of them notices only half of the alterations we make to a record, while the two that name the fault most often cannot tell what a harness left out. With a second writer, the record can be checked. Our sandbox keeps a ledger of what passes between harness and model, and reading the ledger against the record in both directions shows what the record leaves out and what it adds. We test this as an auditor would use it, on one unseen record at a time. On runs driven by a third vendor's model, with every rule committed before the records existed, it reports 64 of 66 alterations and flags none of 84 clean records. On held-out records in five formats it reports 109 of 115. What makes a record evidence is a second writer, not a better reader.

查看此版本解读 →查看此版本原文 ↗