仅凭输出审查无法验证什么：面向研究智能体的研究契约

View PDF HTML (experimental)

Abstract:Some defects in an AI-generated study can be identified from its artifacts; others require knowledge of what was approved before execution. We propose study contracts that bind declared experimental choices, run obligations and claim scope to recorded execution evidence, and distinguish this contract-relative verification from scientific truth. A diagnostic using eight self-authored clean/mutated pairs illustrates the information boundary. A deterministic checker applying a registered, fault-specific rule to approved and executed objects detected all eight registered mutations. Across eighteen recorded judge aliases given individual metadata-filtered packages without pair context or the registry-selected fault label, 104 of 144 mutated evaluation cases received defect flags; the remaining cases comprised 32 abstentions and eight terminal failures, with no explicit clean decisions on mutated cases. Some packages retained approval and execution fields, including digests. The prompt instructed judges to abstain when evidence was insufficient. These results characterize a deliberately information-asymmetric development setting; they do not isolate the effect of authoritative information from differences in task specification and rule selection, and they are not comparative verifier quality or agent reward hacking. We identify full-information comparisons, legitimate-adaptation controls and closed-loop agent evaluations as necessary tests of whether contract checks improve useful compliant completion under optimization.

Submission history

From: Eitan Waks [view email]
[v1] Thu, 8 Oct 2026 11:42:23 UTC (30 KB)