List your company

Visual Color Reasoning on ColorBench cleaned

57.87ScoreJigsaw + CARE
45.535648.737851.9455.1422Dec 16, 2025

Evaluation results

MethodLinks
Jigsaw + CARETraining Environment=Jigsaw puzzles, CARE=true, Backbone=Qwen 2.5 VL 7B, Evaluation Mode=thinking mode (think-answer)2025.1257.87
Mix + CL + CARETraining Environment=Mixed (Jigsaw, PatchFit, Rotation), Curriculum Learning=true, CARE=true, Backbone=Qwen 2.5 VL 7B, Evaluation Mode=thinking mode (think-answer)2025.1257.46
Visual JigsawBackbone=Qwen 2.5 VL 7B, Evaluation Mode=thinking mode (think-answer)2025.1255.21
Jigsaw + CL + CARETraining Environment=Jigsaw puzzles, Curriculum Learning=true, CARE=true, Backbone=Qwen 2.5 VL 7B, Evaluation Mode=thinking mode (think-answer)2025.1254.6
VisualSphinxBackbone=Qwen 2.5 VL 7B, Evaluation Mode=thinking mode (think-answer)2025.1253.99
ViCritBackbone=Qwen 2.5 VL 7B, Evaluation Mode=thinking mode (think-answer)2025.1252.35
Qwen2.5 VLBackbone=Qwen 2.5 VL 7B, Evaluation Mode=thinking mode (think-answer)2025.1251.12
JigsawTraining Environment=Jigsaw puzzles, Backbone=Qwen 2.5 VL 7B, Evaluation Mode=thinking mode (think-answer)2025.1250.31
GRPO-CAREBackbone=Qwen 2.5 VL 7B, Evaluation Mode=thinking mode (think-answer)2025.1248.06
Vision-ZeroBackbone=Qwen 2.5 VL 7B, Evaluation Mode=thinking mode (think-answer)2025.1246.01