Robot Manipulation on LIBERO (Specific Success Rate Metrics)
83.690th Percentile Success RateDINOv2
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| DINOv2Image Encoder=DINOv2 [39], Language Encoder=CLIP [42], Setting=LoRA Fine-tuned2026.05 | 83.6 | 77.5 | 82 | 81 | 67.5 | 78.3 | |
| SigLIPImage Encoder=SigLIP [57], Language Encoder=SigLIP [57], Setting=LoRA Fine-tuned2026.05 | 82.6 | 80.5 | 82 | 74 | 76.5 | 79.1 | |
| CLIPImage Encoder=CLIP [42], Language Encoder=CLIP [42], Setting=LoRA Fine-tuned2026.05 | 78.1 | 79.5 | 79 | 75.5 | 68.5 | 76.1 | |
| DynaFLIPImage Encoder=DynaFLIP (Ours), Language Encoder=DynaFLIP (Ours), Setting=LoRA Fine-tuned2026.05 | 78.1 | 84.5 | 83.5 | 78.5 | 80.5 | 81 | |
| LIVImage Encoder=LIV [34], Language Encoder=LIV [34], Setting=LoRA Fine-tuned2026.05 | 72.7 | 78.5 | 49 | 75.5 | 62 | 67.5 | |
| VC-1Image Encoder=VC-1 [36], Language Encoder=CLIP [42], Setting=LoRA Fine-tuned2026.05 | 72.4 | 83 | 83.5 | 71 | 62 | 74.4 | |
| R3MImage Encoder=R3M [37], Language Encoder=CLIP [42], Setting=LoRA Fine-tuned2026.05 | 38.5 | 67 | 2.5 | 56.5 | 37.5 | 40.4 | |
| DynaFLIPImage Encoder=DynaFLIP (Ours), Language Encoder=DynaFLIP (Ours), Setting=Frozen2026.05 | 31.7 | 70.5 | 37.5 | 51.5 | 16.5 | 41.5 | |
| R3MImage Encoder=R3M [37], Language Encoder=CLIP [42], Setting=Frozen2026.05 | 24.4 | 45 | 0.5 | 53 | 13.5 | 27.3 | |
| SigLIPImage Encoder=SigLIP [57], Language Encoder=SigLIP [57], Setting=Frozen2026.05 | 24.3 | 54.5 | 13 | 52 | 8.5 | 30.5 | |
| LIVImage Encoder=LIV [34], Language Encoder=LIV [34], Setting=Frozen2026.05 | 22.3 | 64 | 6.5 | 51 | 9 | 30.6 | |
| DINOv2Image Encoder=DINOv2 [39], Language Encoder=CLIP [42], Setting=Frozen2026.05 | 14.4 | 75 | 33.5 | 42.5 | 20.5 | 37.2 | |
| CLIPImage Encoder=CLIP [42], Language Encoder=CLIP [42], Setting=Frozen2026.05 | 13.8 | 38.5 | 1.5 | 50 | 9.5 | 22.7 | |
| VC-1Image Encoder=VC-1 [36], Language Encoder=CLIP [42], Setting=Frozen2026.05 | 12.8 | 52.5 | 11.5 | 52 | 12.5 | 28.3 |