← 피드로
[Submitted on 21 Aug 2025 (v1), last revised 7 Sep 2026 (this version, v2)]
Abstract:Tool use is often assumed to monotonically improve reasoning, where external evidence is expected to help when relevant and be ignored when irrelevant. We show that this assumption fails in a state-dependent way. Across benchmarks with Python and Wikipedia tools, external evidence reliably helps when initial beliefs are weak, but can flip already-correct answers when those beliefs are strong. We frame this as a misallocation of revision authority, arguing that deferring to external evidence is suboptimal when internal support for the correct answer surpasses the tool’s expected output quality. This predicts that harm should concentrate on high-confidence no-tool cases. We test this prediction with threshold localization, wrong-trace audits, and a same-clue intervention showing that revision framing changes the damage caused by misleading evidence. These findings suggest that mixed no-tool/tool-assisted inference should arbitrate authority rather than privilege tool evidence by default. As a minimal demonstration, we introduce CASE, a label-free controller that selects between no-tool and tool-assisted trajectories using answer-state certainty and improves over existing confidence-based arbitration this http URL code for our experiments is available at this https URL.
Submission history
From: Junnan Liu [view email]
[v1]
Thu, 21 Aug 2025 17:50:24 UTC (1,765 KB)
[v2]
Mon, 7 Sep 2026 13:32:46 UTC (610 KB)
추출 본문 · 출처: arxiv.org · https://arxiv.org/abs/2508.15754