ICLR 2026 论文提出带可证明保证的自动化电路发现算法
burny_tech · x · 2026-10-21
The arXiv paper "Formal Mechanistic Interpretability: Automated Circuit Discovery with Provable Guarantees" (Hadad, Katz, Bassan; to appear at ICLR 2026) leverages recent advances in neural network verification to propose automated algorithms that discover circuits with provable guarantees.
Three types of guarantees:
- Input domain robustness: the circuit agrees with the model across a continuous input region;
- Robust patching: certifies circuit alignment under continuous patching perturbations;
- Minimality: formalizes a wide array of notions of succinctness.
The authors uncover novel theoretical connections among the three guarantee families, with critical implications for algorithm convergence. Experiments with state-of-the-art verifiers on various vision models show substantially stronger robustness guarantees than standard circuit discovery methods, establishing a principled foundation for provable circuit discovery.
所属事件:ICLR 2026 论文：带可证明保证的自动化电路发现算法(2 条相关)→
「安全」频道最新
- PHP 核心现 CVSS 满分 RCE 漏洞，10 月 21 日发紧急补丁 — cyb3rops · 2026-10-21
- Anaconda 推 MCP 帮编码 Agent 核验依赖包,防包名幻觉供应链攻击 — anacondainc · 2026-10-21
- NIH 资助论文免付费公开：作者手稿须无禁期上传 PMC — anshulkundaje · 2026-10-21
- Creed 论文：从代码推断开发者信念，做首个语义级安全审计 — mboehme_ · 2026-10-21
- Spotify 年内下架 5000 万首 AI 垃圾音乐，力保人类创作者 — moustaki · 2026-10-21
- METR 认可 Anthropic 风险报告结论,但外部专家批其风险估计方法粗糙 — DKokotajlo · 2026-10-21