Robotic Manipulation on CSOT-Bench Ours (fine-tuning)
83.1Scene Success RateTriRelVLA
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| TriRelVLARepresentation Category=Structured, Evaluation Protocol=Fine-tuned2026.05 | 83.1 | 91.5 | 80.3 | 84.9 | |
| SemanticVLARepresentation Category=Structured, Evaluation Protocol=Fine-tuned2026.05 | 82.7 | 91.3 | 78.5 | 84.2 | |
| CogVLARepresentation Category=Structured, Evaluation Protocol=Fine-tuned2026.05 | 81.2 | 88.1 | 78.4 | 82.5 | |
| SpatialVLARepresentation Category=Implicit, Evaluation Protocol=Fine-tuned2026.05 | 79.5 | 87.6 | 75.4 | 80.8 | |
| CogACTRepresentation Category=Implicit, Evaluation Protocol=Fine-tuned2026.05 | 76.1 | 84.5 | 73.8 | 78.1 | |
| OpenVLARepresentation Category=Implicit, Evaluation Protocol=Fine-tuned2026.05 | 74.6 | 82.4 | 72 | 76.3 | |
| DiffusionPolicyRepresentation Category=Implicit, Evaluation Protocol=Fine-tuned2026.05 | 68.7 | 80.2 | 65.3 | 71.4 | |
| OctoRepresentation Category=Implicit, Evaluation Protocol=Fine-tuned2026.05 | 66.3 | 78 | 71.5 | 72 |