Q4 Compliance Audit Review - Jira & SlackCompliance & Regulatory
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.
100%
Claude Opus 4.8
100%
Gemini 3.1 Pro
97%
GPT-5.5
72%
Gemini 3.5 Flash
Extended metrics
Median Time
Wall-clock minutes, median across trials. N = 10 trials per model.
7m
Gemini 3.5 Flash
3m
Claude Opus 4.8
2m
Gemini 3.1 Pro
2m
GPT-5.5
Tool Calls
Tool invocations, median across trials — counted from the trajectories. N = 10 trials per model.
59
Gemini 3.5 Flash
40
Gemini 3.1 Pro
38
Claude Opus 4.8
37
GPT-5.5
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 (19) | ||||
| answer_audit_items_json | 100% | 100% | 100% | 80% |
| answer_in_progress_items_json | 100% | 100% | 100% | 80% |
| answer_piermont_count_and_ranking_json | 60% | 100% | 100% | 20% |
| comment_on_every_in_progress_item_in_database | 100% | 100% | 100% | 70% |
| commented_on_every_in_progress_item | 100% | 100% | 100% | 70% |
| comments_only_on_in_progress_items | 100% | 100% | 100% | 100% |
| posted_slack_message | 100% | 100% | 100% | 30% |
| retrieved_banking_channel_history | 100% | 100% | 100% | 100% |
| retrieved_bo1_nacha_ach_audit | 100% | 100% | 100% | 100% |
| retrieved_bo2_pm_dispute_audit | 100% | 100% | 100% | 100% |
| retrieved_bo33_kyc_audit | 100% | 100% | 100% | 100% |
| retrieved_every_in_progress_item | 100% | 100% | 100% | 100% |
| slack_message_in_database | 100% | 100% | 100% | 30% |
| slack_message_to_correct_channel | 100% | 100% | 100% | 30% |
| answer_describes_the_top_posters | 100% | 100% | 100% | 70% |
| answer_states_audit_conclusions | 100% | 100% | 100% | 70% |
| enumerated_bank_operations_project | 100% | 100% | 100% | 100% |
| resolved_top_poster_identities | 100% | 100% | 100% | 40% |
| retrieved_bo1_comments | 100% | 100% | 100% | 100% |
Failure modes on this task
13 tagged verifier failures, by the stage they were assigned to. Higher is worse.
| Failure stage | GPT-5.5 | Claude Opus 4.8 | Gemini 3.1 Pro | Gemini 3.5 Flash |
|---|---|---|---|---|
| Answer2 modes | 4 | — | — | 9 |