List your company

Decision classification on Damage-audit 139-case extended (held-out)

84.2Action AccuracyB1b
73.69676.42379.1581.877Jun 9, 2026

Evaluation results

MethodLinks
B1bExecutor regime=Haiku 4.52026.0684.237.4
B1bExecutor regime=Opus 4.62026.0682.736.7
Auto-EISRExecutor regime=Haiku 4.5, seeds=3, regression gate=disabled2026.0681.350.1
Auto-EISRExecutor regime=Opus 4.6, seeds=3, regression gate=disabled2026.0680.650.4
Human-v8Executor regime=Opus 4.62026.0677.248
v1Executor regime=Haiku 4.52026.067735.3
Human-v8Executor regime=Haiku 4.52026.067550
v1Executor regime=Opus 4.62026.0674.136.7