Pairwise progress-judgment on RoboPulse Overall
83Overall Average AccuracyRobo-Dopamine
Evaluation results
| Method | Links | |
|---|---|---|
| Robo-Dopaminezero-shot=true2026.03 | 83 | |
| VLACzero-shot=true2026.03 | 71 | |
| GVLzero-shot=true2026.03 | 71 | |
| Gemini 3 Pro Previewzero-shot=true2026.03 | 66 | |
| CLIP ViT-L/14 (I2I)zero-shot=true2026.03 | 59 | |
| Qwen3-VL-8B-Instructzero-shot=true2026.03 | 59 | |
| CLIP ViT-B/32 (I2I)zero-shot=true2026.03 | 57 | |
| Qwen3-VL-4B-Instructzero-shot=true2026.03 | 56 | |
| GPT-5.2zero-shot=true2026.03 | 53 | |
| CLIP ViT-B/32 (T2I)zero-shot=true2026.03 | 47 | |
| CLIP ViT-L/14 (T2I)zero-shot=true2026.03 | 46 |