← 研踪 PaperTrailarXiv:2611.20459 · v1
Designing Deceptive Model Organisms that Successfully Evade Auditing
首次提交 2026-11-16 11:27
版本更新 2026-11-16 11:27
雷达观测 2026-11-17 13:07
摘要
AI models can exhibit deceptive behaviors such as strategically underperforming on evaluations (sandbagging), faking alignment, or covertly acting on behalf of a hidden principal. Safe deployment therefore requires auditing techniques that can reliably detect such behaviors. To develop and test these techniques, we can train model organisms (MOs): models that exhibit deceptive behaviors and serve as realistic evaluation testbeds. However, for such testbeds to be informative, MOs must actively anticipate and resist auditing---a gap in prior work that can lead to overestimation of auditing effectiveness and underestimation of safety risks. We address this gap by designing MOs that capture worst-case deceptive behavior and evade high-effort auditing. Focusing on sandbagging, we frame MO design as a security game between an auditor and a deceptive model and identify three design requirements for sandbagging MOs: (i) effectiveness (underperforming on targeted capabilities), (ii) utility preservation (maintaining performance on all other capabilities), and (iii) robustness (resisting auditing). We propose three MOs: two optimized against blackbox auditing techniques and one optimized against both blackbox and whitebox techniques. Through a comprehensive evaluation, we show that, unlike MOs from prior work, our MOs satisfy all three requirements, providing worst-case baselines for developing and evaluating future auditing techniques.
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 · 2026-11-16 11:27 · 最新观测版本
Designing Deceptive Model Organisms that Successfully Evade Auditing
AI models can exhibit deceptive behaviors such as strategically underperforming on evaluations (sandbagging), faking alignment, or covertly acting on behalf of a hidden principal. Safe deployment therefore requires auditing techniques that can reliably detect such behaviors. To develop and test these techniques, we can train model organisms (MOs): models that exhibit deceptive behaviors and serve as realistic evaluation testbeds. However, for such testbeds to be informative, MOs must actively anticipate and resist auditing---a gap in prior work that can lead to overestimation of auditing effectiveness and underestimation of safety risks. We address this gap by designing MOs that capture worst-case deceptive behavior and evade high-effort auditing. Focusing on sandbagging, we frame MO design as a security game between an auditor and a deceptive model and identify three design requirements for sandbagging MOs: (i) effectiveness (underperforming on targeted capabilities), (ii) utility preservation (maintaining performance on all other capabilities), and (iii) robustness (resisting auditing). We propose three MOs: two optimized against blackbox auditing techniques and one optimized against both blackbox and whitebox techniques. Through a comprehensive evaluation, we show that, unlike MOs from prior work, our MOs satisfy all three requirements, providing worst-case baselines for developing and evaluating future auditing techniques.
查看此版本解读 →查看此版本原文 ↗