data/evaluation_results.csv

model,category,accuracy,precision,recall,num_samples
Qwen2.5-VL-32B-BASE,masterCategory,0.909,0.9196051103650724,0.909,1000
Qwen2.5-VL-32B-BASE,gender,0.546,0.9259626959624715,0.546,1000
Qwen2.5-VL-32B-BASE,subCategory,0.432,0.7070035848765855,0.432,1000
Qwen2.5-VL-32B-SFT,masterCategory,0.7898550724637681,0.8517880287498527,0.7898550724637681,690
Qwen2.5-VL-32B-SFT,gender,0.9,0.8784316770186336,0.9,690
Qwen2.5-VL-32B-SFT,subCategory,0.9173913043478261,0.9378322745486548,0.9173913043478261,690
Qwen2-VL-72B-BASE,masterCategory,0.968968968968969,0.9711267688093789,0.968968968968969,999
Qwen2-VL-72B-BASE,gender,0.7607607607607607,0.9354341592843324,0.7607607607607607,999
Qwen2-VL-72B-BASE,subCategory,0.34134134134134136,0.6784829173652965,0.34134134134134136,999
Qwen2-VL-72B-SFT,masterCategory,0.993993993993994,0.9940108529582213,0.993993993993994,999
Qwen2-VL-72B-SFT,gender,0.9169169169169169,0.9144956029794004,0.9169169169169169,999
Qwen2-VL-72B-SFT,subCategory,0.9419419419419419,0.9512743495222181,0.9419419419419419,999
GPT-5-Mini,masterCategory,0.981,0.9810138759482104,0.981,1000
GPT-5-Mini,gender,0.907,0.9260515702929443,0.907,1000
GPT-5-Mini,subCategory,0.897,0.944355065421394,0.897,1000