Model openaio3

Training model: o3
Inference model: gpt-4o-mini

Investigations

Performance

DatasetValidation accuracyTest accuracy
espionage0.950.8
potions0.80.75
southgermancredit0.59615384615384610.49557522123893805
timetravel_insurance0.80.7
titanic0.8039215686274510.6862745098039216
wisconsin0.93650793650793650.8333333333333334