Model gemini10

Training model: gemini-2.0-flash
Inference model: gpt-4o-mini

Investigations

Performance

DatasetValidation accuracyTest accuracy
espionage0.950.8
potions0.750.6
southgermancredit0.45192307692307690.46017699115044247
timetravel_insurance0.750.8
titanic0.66666666666666660.5686274509803921
wisconsin0.71428571428571430.6515151515151515