Frontier Log

Benchmark · Active

ExploitGym

Created by Berkeley RDI, 11 May 2026 · Higher is better · Website

Settings

Task setScoringMitigationsReported by
ExploitGym 1.0 (869 tasks)ScoreMitigations offOpenAIGoogle
ExploitGym 1.0 (869 tasks)ScoreMitigations onOpenAI
ExploitGym 1.0 (869 tasks)ScoreNot specifiedOpenAIMetaDeepSeekMoonshot AI
ExploitGym 1.0 (869 tasks)Successful exploitsMitigations offAnthropicMetaZ.ai
ExploitGym 1.0 (869 tasks)Successful exploitsNot specifiedMeta

Bold: Benchmark default based on Paper and Leaderboard · Download image

Results

0200400600869Jul 2026Oct 202620Muse Wonder Full5Muse Spark 1.10200400600869Jul 2026Oct 202620Muse Wonder Full5Muse Spark 1.1
ModelHarness and effortScoreSource
Muse Wonder Full12 Oct 2026Harness not specifiedEffort not specified20 of 869Muse Wonder model cardChecked A person checked this number against its source on 20 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.intended flags only, after a harness correction that added the intended vulnerability to the prompt
Muse Spark 1.19 Jul 2026Harness not specifiedEffort not specified5 of 869Muse Spark 1.1 Evaluation ReportChecked A person checked this number against its source on 20 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.pass@1 · 2h timeout

All benchmarks