model,task,accuracy,correct,total Qwen2.5-VL-3B-Instruct-judge-enc,depth_estimation,0.6118,208,340 Qwen2.5-VL-3B-Instruct-judge-enc,generation_controllable,0.5816,196,337 Qwen2.5-VL-3B-Instruct-judge-enc,generation_editing,0.55,187,340 Qwen2.5-VL-3B-Instruct-judge-enc,generation_inpainting_high_level,0.6147,209,340 Qwen2.5-VL-3B-Instruct-judge-enc,generation_inpainting_low_level,0.6029,205,340 Qwen2.5-VL-3B-Instruct-judge-enc,generation_t2i,0.6912,235,340 Qwen2.5-VL-3B-Instruct-judge-enc,instance_segmentation,0.6161,626,1016 Qwen2.5-VL-3B-Instruct-judge-enc,keypoint,0.6529,222,340 Qwen2.5-VL-3B-Instruct-judge-enc,lowlevel-deblur,0.7818,172,220 Qwen2.5-VL-3B-Instruct-judge-enc,lowlevel-derain,0.8765,298,340 Qwen2.5-VL-3B-Instruct-judge-enc,lowlevel-desnow,0.9149,301,329 Qwen2.5-VL-3B-Instruct-judge-enc,lowlevel-super-resolution,0.5441,185,340 Qwen2.5-VL-3B-Instruct-judge-enc,object_detection,0.7422,757,1020 Qwen2.5-VL-3B-Instruct-judge-enc,referring_segmentation,0.5833,595,1020 Qwen2.5-VL-3B-Instruct-judge-enc,semantic_segmentation,0.6079,727,1196 Qwen2.5-VL-7B-Instruct-judge-enc,depth_estimation,0.7176,244,340 Qwen2.5-VL-7B-Instruct-judge-enc,generation_controllable,0.6103,166,272 Qwen2.5-VL-7B-Instruct-judge-enc,generation_editing,0.5605,190,339 Qwen2.5-VL-7B-Instruct-judge-enc,generation_inpainting_high_level,0.6206,211,340 Qwen2.5-VL-7B-Instruct-judge-enc,generation_inpainting_low_level,0.5941,202,340 Qwen2.5-VL-7B-Instruct-judge-enc,generation_t2i,0.6735,229,340 Qwen2.5-VL-7B-Instruct-judge-enc,instance_segmentation,0.6248,631,1010 Qwen2.5-VL-7B-Instruct-judge-enc,keypoint,0.6706,228,340 Qwen2.5-VL-7B-Instruct-judge-enc,lowlevel-deblur,0.7318,161,220 Qwen2.5-VL-7B-Instruct-judge-enc,lowlevel-derain,0.9059,308,340 Qwen2.5-VL-7B-Instruct-judge-enc,lowlevel-desnow,0.9331,307,329 Qwen2.5-VL-7B-Instruct-judge-enc,lowlevel-super-resolution,0.5971,203,340 Qwen2.5-VL-7B-Instruct-judge-enc,object_detection,0.7529,768,1020 Qwen2.5-VL-7B-Instruct-judge-enc,referring_segmentation,0.6206,633,1020 Qwen2.5-VL-7B-Instruct-judge-enc,semantic_segmentation,0.6279,751,1196