arXiv cs.CL· Xiaochen Chang·· 11 小时前AI 评分38
隔离异议的 harness 如何减少智能体写作中的过度防御性表述
A Harness Reducing over-Defensive Writing in Agents by Isolating Objections
AI 导读
研究者提出一种将审稿、过滤与写作分离的 harness，写作智能体只接收过滤器筛选出的问题，从而减少智能体扩写报告时的过度防御性写作。在 25 份商业项目草稿扩写为正式报告的评测中，防御性命题从 59 降至 2，减少 96.61%，未发现无依据的事实添加，来源主张保留率从 80.35% 升至 93.12%。
正文 · 原文
Abstract:When expanding a report, writing with agents may introduce a stronger claim than the article makes and then disclaim it. This is called over-defensive writing. We assume that this is because the agent is writing for responding to an adversarial AI reviewer, instead of a human reader. The defensive context may also influnce the consequent when we simply ban defensive writing in the output through prompts. In order to reduce over-defensive writing, we produce a harness that separates the reviewer, filter, and writer. The writer agent receives only the problems choosen by the filter, which is an independent context. We evaluate this harness on 25 business-project drafts expanded into formal reports, the result shows that the defensive propositions decreased from 59 to 2, which is a 96.61\% reduction, no unsupported factual additions were identified while the source-claim retention rises from 80.35% to 93.12%.
| Comments: | defensive writing; context isolation; harness |
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2610.36717 [cs.CL] |
| (or arXiv:2610.36717v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2610.36717 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Xiaochen Chang [view email]
[v1]
Tue, 27 Oct 2026 06:32:13 UTC (16 KB)
来源：arXiv cs.CL · arxiv.org