Audit card authorizations after account lock timestampRisk & Fraud
How each model performed on this task. Every chart states its N, and the reward carries a 95% confidence interval once there is more than one trial to measure.
Core metrics
Check Pass %
Mean share of the task's checks passing, averaged over trials. N = 10 trials per model.
86%
GPT-5.5
79%
Claude Opus 4.8
46%
Gemini 3.5 Flash
36%
Gemini 3.1 Pro
Extended metrics
Median Time
Wall-clock minutes, median across trials. N = 10 trials per model.
19m
Gemini 3.1 Pro
14m
GPT-5.5
7m
Claude Opus 4.8
5m
Gemini 3.5 Flash
Tool Calls
Tool invocations, median across trials — counted from the trajectories. N = 10 trials per model.
250
Gemini 3.1 Pro
248
GPT-5.5
91
Claude Opus 4.8
41
Gemini 3.5 Flash
Verifier × model pass rate
How often each verifier passed, per model. Green is passing, red is failing — a red band across a row is a verifier nothing clears.
| Verifier | GPT-5.5n=10 | Claude Opus 4.8n=10 | Gemini 3.1 Pron=10 | Gemini 3.5 Flashn=10 |
|---|---|---|---|---|
| checks (11) | ||||
| gold_leaked_dollars | 90% | 80% | 30% | 50% |
| gold_offending_accounts | 90% | 80% | 30% | 60% |
| gold_offending_auths | 90% | 80% | 30% | 60% |
| gold_worst_account_dollars | 90% | 100% | 100% | 50% |
| rubric_root_cause | 90% | 60% | 20% | 10% |
| rubric_uses_all_connectors | 60% | 80% | 10% | 20% |
| scopes_lock_leak_correctly | 90% | 70% | 30% | 70% |
| gold_accounts_multi_auth | 90% | 80% | 40% | 60% |
| gold_cards_never_propagated | 90% | 60% | 40% | 30% |
| process_floor | 100% | 100% | 100% | 100% |
| rubric_root_cause_ties_to_leaks | 90% | 100% | 100% | 80% |