智能体 AI 中的验证与自我改进：基础与边界
Verification and Self-Improvement in Agentic AI: Foundations and Limits
AI 导读
一篇 arXiv 论文提出区分自改进智能体的搜索增益与已接受任务覆盖扩展的框架，引入支持与转录接口、原生可达集与闭包边界等概念，并用语义包含准则和 guard-transfer 不变性约束验证逃逸。
正文 · 原文
Abstract:A self-improving agent can find accepted solutions more often without expanding the tasks its verification interface can accept. We extend task-relative verification with explicit support and transcript interfaces to distinguish search gains from changes in accepted-task coverage. A stage's native reach contains tasks accepted with default support; its closure frontier contains tasks accepted with any support the declared interface permits. A semantic containment criterion identifies the authority or exposure condition that every verified escape must violate, while guard-transfer invariance prevents a change of representation from being counted as a capability expansion. An executable Boolean-program repair agent makes the distinction concrete: over 36,864 paired runs, counterexample feedback reduces verifier calls without changing coverage, whereas an enforced permission change admits a new target with a proof excluding every old-interface program. Exact and uniformly gapped randomized verification share this account; independent majority amplification improves reliability while preserving the accepted language. Separate adaptive-selection contracts state what is additionally required to interpret a proposed successor's score as evidence of deployment improvement. Classical hierarchy results and finite audits provide supporting boundaries. The framework specifies the evidence needed to establish improved search, wider accepted capability, and reliable successor selection.
| Comments: | 36 pages, 7 figures, 4 tables. Revised around search gains, accepted-task coverage, and successor selection; added executable Boolean-program repair experiments and structural analysis; detailed proofs in appendices. Includes reproducible research artifacts and partial Lean formalization with explicit open obligations |
| Subjects: | Artificial Intelligence (cs.AI); Computational Complexity (cs.CC) |
| Cite as: | arXiv:2610.10611 [cs.AI] |
| (or arXiv:2610.10611v2 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2610.10611 arXiv-issued DOI via DataCite |
Submission history
From: Chien-Ping Lu [view email]
[v1]
Wed, 7 Oct 2026 06:47:49 UTC (90 KB)
[v2]
Fri, 30 Oct 2026 14:29:44 UTC (603 KB)
来源：arXiv cs.AI · arxiv.org