Reconstruct cancelled-check return-then-re-debit ledger and net balance impactReconciliation & Settlement
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.
71%
Gemini 3.1 Pro
62%
Claude Opus 4.8
27%
GPT-5.5
0%
Gemini 3.5 Flash
Extended metrics
Median Time
Wall-clock minutes, median across trials. N = 10 trials per model.
23m
GPT-5.5
8m
Gemini 3.1 Pro
6m
Claude Opus 4.8
4m
Gemini 3.5 Flash
Tool Calls
Tool invocations, median across trials — counted from the trajectories. N = 10 trials per model.
226
GPT-5.5
64
Gemini 3.1 Pro
37
Claude Opus 4.8
23
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 (15) | ||||
| cp_fully_corrected_count | 70% | 80% | 90% | 0% |
| cp_gross_dollars_moved | 0% | 40% | 70% | 0% |
| cp_net_exposure_usd | 0% | 40% | 70% | 0% |
| cp_outstanding_overdebit_usd | 0% | 40% | 70% | 0% |
| cp_outstanding_undercredit_usd | 0% | 40% | 70% | 0% |
| cp_over_corrected_count | 10% | 60% | 70% | 0% |
| cp_pattern_check_count | 60% | 100% | 90% | 0% |
| cp_under_corrected_count | 10% | 60% | 70% | 0% |
| cp_window_end | 70% | 100% | 90% | 0% |
| cp_window_start | 70% | 100% | 90% | 0% |
| reconstruction_method | 10% | 20% | 0% | 0% |
| cohort_and_remediation | 30% | 0% | 0% | 0% |
| posted_deliverable_to_slack | 80% | 100% | 70% | 0% |
| process_floor_sql | 100% | 100% | 100% | 20% |
| root_cause_and_control | 80% | 100% | 70% | 10% |