Share your thoughts, 1 month free Claude Pro on usSee more

Sabotage Detection on ImpossibleBench

100Average Precision

Monitor

Updated 4d ago

Evaluation Results

Method	Links
Monitor 2026.04		100
Bayesian 2026.04		100
Bayesian 2026.04		100
Buffer 2026.04		99
Buffer 2026.04		99
Bayesian 2026.04		96.9
Bayesian 2026.04		96.9
Buffer 2026.04		96.5
Buffer 2026.04		96.5
Meerkat 2026.04		96.2
Naive Agent 2026.04		94.2
Monitor 2026.04		94
Meerkat 2026.04		92.7
Bayesian 2026.04		92.6
Bayesian 2026.04		92.5
Meerkat 2026.04		92
Buffer 2026.04		90
Buffer 2026.04		90
Bayesian 2026.04		88.2
Bayesian 2026.04		88.1
Meerkat 2026.04		86.9
Buffer 2026.04		86.7
Buffer 2026.04		86.7
Meerkat 2026.04		85.2
Monitor 2026.04		84.8
Monitor 2026.04		83.4
Meerkat 2026.04		83.3
Naive Agent 2026.04		82.4
Monitor 2026.04		80
Naive Agent 2026.04		78.9
Monitor 2026.04		78.6
Monitor 2026.04		78
Naive Agent 2026.04		77.7
Meerkat 2026.04		74.9
Monitor 2026.04		74.5
Meerkat 2026.04		73.6
Naive Agent 2026.04		68.7
Naive Agent 2026.04		62.6
Naive Agent 2026.04		50
Naive Agent 2026.04		50