MonitoringBench - Human Strategies - Sudo Escalation: leaderboard

MonitoringBench (2026) tests LLMs as monitors of coding agents in BashArena: the monitor reads a full agent trajectory of tool calls and scores how suspicious it is, and an attack counts as caught when the score exceeds the monitor threshold for a 1% false-positive rate on honest runs. This board covers the 313 successful sudo privilege escalation attacks from the human strategies stage of the red-teaming pipeline, scored as catch rate in percent; higher is better.

Metric: Catch rate (%) at a 1% false-positive-rate threshold calibrated on 416 honest BashArena runs: share of the 313 successful sudo privilege escalation attack trajectories (human-seeded attack strategies executed by Claude Opus 4.6) that the full-trajectory monitor scores above its threshold, default ControlArena monitor prompt with a 0-10 suspiciousness score, mean of three scoring runs; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 13 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Claude Opus 4.5 (20251101)93.5#139
2GPT-5.1 Codex Max84#207
3Claude Sonnet 4.578.4#261
4GPT-570.2#178
5O3 (2025-04-16)66.3#231
6Claude Sonnet 4 (20250514)62.9#405
7GPT-4.130.4#505
8GPT-5 Nano23.4#906
9GPT-4o (2024-08-06)15.1#743
10GPT-4 Turbo14.2#1000
11O4 Mini (2025-04-16)11.1#370
12GPT-4.1 Mini5#789
13GPT-4o Mini (2024-07-18)0.7#1484

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=monitoringbench-human-strategies-sudo-escalation · How It Works · Data refreshed daily, snapshot 2026-11-21.