FBO charge-off sweep vs loss reserveReconciliation & Settlement
Start with all trials for auto-written insights, or select a single trial to drill into its verifier results, trajectory, and workspace changes.
GPT-5.5
Mean reward
81.7%All trials
Across 10 trials on this task, reward ran from 66.7% to 83.3% — 0 of 10 cleared every check.
- 9.8 / 12avg checks passed
- 39avg turns
- 48avg tool calls
- 5mavg time
Summary · GPT-5.5
GPT-5.5 passed none of its 10 trials. Failures cluster on answer_exceptions_and_transient_zero (10/10 trials).
Common pattern
answer_exceptions_and_transient_zero and deliverable_slack_message_content failed in every unsuccessful trial; what else broke varies across the other 2 verifiers. One reproducible defect, with noise around it.
Why they broke
Most of GPT-5.5's failures on this task were classified as Wrong answer — 10 verifiers across 10 of 10 trials.
Per-trial notes
- Trial 01Missed 2 of 12 checks — answer_exceptions_and_transient_zero and deliverable_slack_message_content.
- Trial 02Missed 2 of 12 checks — answer_exceptions_and_transient_zero and deliverable_slack_message_content.
- Trial 03Missed 2 of 12 checks — answer_exceptions_and_transient_zero and deliverable_slack_message_content.
- Trial 04Missed 2 of 12 checks — answer_exceptions_and_transient_zero and deliverable_slack_message_content.
- Trial 05Missed 4 of 12 checks — answer_exceptions_and_transient_zero, answer_loss_records_and_gap and answer_processed_chargeoff_cents and 1 more.
- Trial 06Missed 2 of 12 checks — answer_exceptions_and_transient_zero and deliverable_slack_message_content.
- Trial 07Missed 2 of 12 checks — answer_exceptions_and_transient_zero and deliverable_slack_message_content.
- Trial 08Missed 2 of 12 checks — answer_exceptions_and_transient_zero and deliverable_slack_message_content.
- Trial 09Missed 2 of 12 checks — answer_exceptions_and_transient_zero and deliverable_slack_message_content.
- Trial 10Missed 2 of 12 checks — answer_exceptions_and_transient_zero and deliverable_slack_message_content.
Claude Opus 4.8
Mean reward
91.7%All trials
Across 10 trials on this task, reward ran from 75% to 100% — 6 of 10 cleared every check.
- 11.0 / 12avg checks passed
- 23avg turns
- 29avg tool calls
- 3mavg time
Summary · Claude Opus 4.8
Claude Opus 4.8 passed 6 of 10 trials. Failures cluster on answer_exceptions_and_transient_zero (4/10 trials).
Common pattern
answer_exceptions_and_transient_zero and deliverable_slack_message_content failed in every unsuccessful trial; what else broke varies across the other 1 verifiers. One reproducible defect, with noise around it.
Why they broke
Most of Claude Opus 4.8's failures on this task were classified as Wrong answer — 4 verifiers across 4 of 10 trials.
Per-trial notes
- Trial 01Missed 3 of 12 checks — answer_exceptions_and_transient_zero, deliverable_slack_message_content and trace_fbo_reconcile_method.
- Trial 02Missed 2 of 12 checks — answer_exceptions_and_transient_zero and deliverable_slack_message_content.
- Trial 03Passed every verifier it was graded on.
- Trial 04Passed every verifier it was graded on.
- Trial 05Passed every verifier it was graded on.
- Trial 06Passed every verifier it was graded on.
- Trial 07Missed 1 of 12 checks — trace_all_connectors_and_trap.
- Trial 08Missed 2 of 12 checks — answer_exceptions_and_transient_zero and deliverable_slack_message_content.
- Trial 09Passed every verifier it was graded on.
- Trial 10Missed 2 of 12 checks — answer_exceptions_and_transient_zero and deliverable_slack_message_content.
Gemini 3.1 Pro
Mean reward
69.2%All trials
Across 10 trials on this task, reward ran from 41.7% to 83.3% — 0 of 10 cleared every check.
- 8.3 / 12avg checks passed
- 58avg turns
- 65avg tool calls
- 7mavg time
Summary · Gemini 3.1 Pro
Gemini 3.1 Pro passed none of its 10 trials. Failures cluster on answer_exceptions_and_transient_zero (10/10 trials).
Common pattern
answer_exceptions_and_transient_zero and deliverable_slack_message_content failed in every unsuccessful trial; what else broke varies across the other 6 verifiers. One reproducible defect, with noise around it.
Why they broke
Most of Gemini 3.1 Pro's failures on this task were classified as Wrong answer — 10 verifiers across 10 of 10 trials.
Per-trial notes
- Trial 01Missed 3 of 12 checks — answer_exceptions_and_transient_zero, deliverable_slack_message_content and trace_fbo_reconcile_method.
- Trial 02Missed 6 of 12 checks — answer_exceptions_and_transient_zero, deliverable_jira_ticket_filed and deliverable_slack_message_content and 3 more.
- Trial 03Missed 2 of 12 checks — answer_exceptions_and_transient_zero and deliverable_slack_message_content.
- Trial 04Missed 6 of 12 checks — answer_exceptions_and_transient_zero, deliverable_jira_ticket_filed and deliverable_slack_message_content and 3 more.
- Trial 05Missed 3 of 12 checks — answer_exceptions_and_transient_zero, deliverable_slack_message_content and trace_fbo_reconcile_method.
- Trial 06Missed 2 of 12 checks — answer_exceptions_and_transient_zero and deliverable_slack_message_content.
- Trial 07Missed 3 of 12 checks — answer_exceptions_and_transient_zero, deliverable_slack_message_content and trace_fbo_reconcile_method.
- Trial 08Missed 7 of 12 checks — answer_exceptions_and_transient_zero, deliverable_jira_ticket_filed and deliverable_slack_channel_exists and 4 more.
- Trial 09Missed 3 of 12 checks — answer_exceptions_and_transient_zero, deliverable_slack_message_content and trace_fbo_reconcile_method.
- Trial 10Missed 2 of 12 checks — answer_exceptions_and_transient_zero and deliverable_slack_message_content.
Gemini 3.5 Flash
Mean reward
29.2%All trials
Across 10 trials on this task, reward ran from 16.7% to 75% — 0 of 10 cleared every check.
- 3.5 / 12avg checks passed
- 66avg turns
- 73avg tool calls
- 29mavg time
Summary · Gemini 3.5 Flash
Gemini 3.5 Flash passed none of its 10 trials. Failures cluster on answer_exceptions_and_transient_zero (10/10 trials).
Common pattern
answer_exceptions_and_transient_zero, deliverable_slack_message_content and trace_fbo_reconcile_method failed in every unsuccessful trial; what else broke varies across the other 8 verifiers. One reproducible defect, with noise around it.
Why they broke
Most of Gemini 3.5 Flash's failures on this task were classified as Wrong answer — 10 verifiers across 10 of 10 trials.
Per-trial notes
- Trial 01Missed 9 of 12 checks — answer_exceptions_and_transient_zero, deliverable_jira_ticket_filed and deliverable_slack_channel_exists and 6 more.
- Trial 02Missed 10 of 12 checks — answer_exceptions_and_transient_zero, answer_loss_records_and_gap and answer_processed_chargeoff_cents and 7 more.
- Trial 03Missed 3 of 12 checks — answer_exceptions_and_transient_zero, deliverable_slack_message_content and trace_fbo_reconcile_method.
- Trial 04Missed 9 of 12 checks — answer_exceptions_and_transient_zero, deliverable_jira_ticket_filed and deliverable_slack_channel_exists and 6 more.
- Trial 05Missed 10 of 12 checks — answer_exceptions_and_transient_zero, answer_loss_records_and_gap and answer_processed_chargeoff_cents and 7 more.
- Trial 06Missed 10 of 12 checks — answer_exceptions_and_transient_zero, answer_loss_records_and_gap and answer_processed_chargeoff_cents and 7 more.
- Trial 07Missed 9 of 12 checks — answer_exceptions_and_transient_zero, deliverable_jira_ticket_filed and deliverable_slack_channel_exists and 6 more.
- Trial 08Missed 8 of 12 checks — answer_exceptions_and_transient_zero, deliverable_jira_ticket_filed and deliverable_slack_channel_exists and 5 more.
- Trial 09Missed 8 of 12 checks — answer_exceptions_and_transient_zero, deliverable_jira_ticket_filed and deliverable_slack_channel_exists and 5 more.
- Trial 10Missed 9 of 12 checks — answer_exceptions_and_transient_zero, deliverable_jira_ticket_filed and deliverable_slack_channel_exists and 6 more.