← 研踪 PaperTrailarXiv:2610.39124 · v1
Clean Sheets and Checkpoints: Evidence Complexity for Certifying Sequential Agent Risk
首次提交 2026-10-29 09:13
版本更新 2026-10-29 09:13
雷达观测 2026-10-30 11:25
摘要
Dangerous-capability evaluations often test an AI agent from prepared states at ordered milestones. A clean sheet records no successes. What it certifies about a campaign of $B$ attempts depends on what a failed attempt erases. We model the task as a chain of $k$ milestones divided into checkpoint segments and ask how many zero audit responses certify, at level $α$, a completion probability of at most $ε$. For segments chosen before the audit, counting possible completion windows gives a certificate for any specified budget. The bound also covers adaptive deployment processes that follow the chain's transitions, provided their success probabilities after every complete history are capped by the corresponding audit success probabilities. The audit does not test this transport assumption. In rare-risk phase interiors, each segment is either rarely completed within the budget (active) or nearly always completed (free). The minimum expected number of zero responses is asymptotic to $\log(1/α)\,(B^a/(a!\,ε))^{1/K_A}$, where the $a$ active segments contain $K_A$ milestones and are those at least as long as a budget-dependent threshold. The certificate attains this constant. Every asymptotically optimal evaluator must divide its expected response count equally among active milestones in the limit, with a vanishing share elsewhere. Where a single segment passes from active to free, each of its milestones receives a smaller limiting share than an active milestone; a crossover-aware certificate attains the optimal constant. For selected segments with known or unqueried prefix steps, counting failed cycles can sharpen the certificate.
作者备注：26 pages, 3 figures, 3 tables
观测到的版本
相邻记录可能跳过中间版本；首次发现 v2 及以上也会标为版本更新。
v1 · 2026-10-29 09:13 · 最新观测版本
Clean Sheets and Checkpoints: Evidence Complexity for Certifying Sequential Agent Risk
Dangerous-capability evaluations often test an AI agent from prepared states at ordered milestones. A clean sheet records no successes. What it certifies about a campaign of $B$ attempts depends on what a failed attempt erases. We model the task as a chain of $k$ milestones divided into checkpoint segments and ask how many zero audit responses certify, at level $α$, a completion probability of at most $ε$. For segments chosen before the audit, counting possible completion windows gives a certificate for any specified budget. The bound also covers adaptive deployment processes that follow the chain's transitions, provided their success probabilities after every complete history are capped by the corresponding audit success probabilities. The audit does not test this transport assumption. In rare-risk phase interiors, each segment is either rarely completed within the budget (active) or nearly always completed (free). The minimum expected number of zero responses is asymptotic to $\log(1/α)\,(B^a/(a!\,ε))^{1/K_A}$, where the $a$ active segments contain $K_A$ milestones and are those at least as long as a budget-dependent threshold. The certificate attains this constant. Every asymptotically optimal evaluator must divide its expected response count equally among active milestones in the limit, with a vanishing share elsewhere. Where a single segment passes from active to free, each of its milestones receives a smaller limiting share than an active milestone; a crossover-aware certificate attains the optimal constant. For selected segments with known or unqueried prefix steps, counting failed cycles can sharpen the certificate.
查看此版本解读 →查看此版本原文 ↗