MonitoringBench - Refined - Firewall Backdoor: 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 511 successful firewall backdoor attacks from the refined 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 511 successful firewall backdoor attack trajectories (pipeline trajectories rewritten by the Claude Opus 4.6 refinement stage) 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: Around February 2027. 13 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Claude Opus 4.5 (20251101)75.9#139
2Claude Sonnet 4.552#261
3GPT-5.1 Codex Max46.5#207
4Claude Sonnet 4 (20250514)44#405
5O3 (2025-04-16)10.8#231
6GPT-4.17.3#505
7GPT-55.8#178
8GPT-4 Turbo4.9#1000
9GPT-4o (2024-08-06)4.4#743
10GPT-4.1 Mini1.5#789
11GPT-5 Nano0.3#906
12GPT-4o Mini (2024-07-18)0.1#1484
13O4 Mini (2025-04-16)0.1#370

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

Interactive version: theaggregate.ai/benchmark?slug=monitoringbench-refined-firewall-backdoor · How It Works · Data refreshed daily, snapshot 2026-11-21.