Multimodal Understanding on SEED-Bench (cleaned)
88.64Overall ScoreJigsaw
Evaluation results
| Method | Links | |
|---|---|---|
| JigsawTraining Environment=Jigsaw puzzles, Backbone=Qwen 2.5 VL 7B, Evaluation Mode=thinking mode (think-answer)2025.12 | 88.64 | |
| GRPO-CAREBackbone=Qwen 2.5 VL 7B, Evaluation Mode=thinking mode (think-answer)2025.12 | 88.58 | |
| 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)2025.12 | 88.31 | |
| ViCritBackbone=Qwen 2.5 VL 7B, Evaluation Mode=thinking mode (think-answer)2025.12 | 88.3 | |
| Vision-ZeroBackbone=Qwen 2.5 VL 7B, Evaluation Mode=thinking mode (think-answer)2025.12 | 87.76 | |
| VisualSphinxBackbone=Qwen 2.5 VL 7B, Evaluation Mode=thinking mode (think-answer)2025.12 | 87.27 | |
| Qwen2.5 VLBackbone=Qwen 2.5 VL 7B, Evaluation Mode=thinking mode (think-answer)2025.12 | 87.15 | |
| Jigsaw + CARETraining Environment=Jigsaw puzzles, CARE=true, Backbone=Qwen 2.5 VL 7B, Evaluation Mode=thinking mode (think-answer)2025.12 | 87.11 | |
| Jigsaw + CL + CARETraining Environment=Jigsaw puzzles, Curriculum Learning=true, CARE=true, Backbone=Qwen 2.5 VL 7B, Evaluation Mode=thinking mode (think-answer)2025.12 | 87.1 | |
| Visual JigsawBackbone=Qwen 2.5 VL 7B, Evaluation Mode=thinking mode (think-answer)2025.12 | 86.22 |