Robot Manipulation Generalization On Gembench
Metrics
Average Success Rate
Average Success Rate (L1)
Average Success Rate (L2)
Average Success Rate (L3)
Average Success Rate (L4)
Results
Performance results of various models on this benchmark
Model Name | Average Success Rate | Average Success Rate (L1) | Average Success Rate (L2) | Average Success Rate (L3) | Average Success Rate (L4) | Paper Title | Repository |
|---|---|---|---|---|---|---|---|
| PolarNet | 38.4 | 77.7±0.9 | 37.1±1.4 | 38.5±1.7 | 0.1±0.2 | PolarNet: 3D Point Clouds for Language-Guided Robotic Manipulation | - |
| 3D-LOTUS++ | 48.0 | 68.7±0.6 | 64.5±0.9 | 41.5±1.8 | 17.4±0.4 | Towards Generalizable Vision-Language Robotic Manipulation: A Benchmark and LLM-guided 3D Policy | - |
| 3D diffuser actor | 44.0 | 91.9±0.8 | 43.4±2.8 | 37.0±2.2 | 0.0±0.0 | 3D Diffuser Actor: Policy Diffusion with 3D Scene Representations | |
| RVT-2 | 44.0 | 89.1±0.8 | 51.0±2.3 | 36.0±2.2 | 0.0±0.0 | RVT-2: Learning Precise Manipulation from Few Demonstrations | - |
| Hiveformer | 30.4 | 60.3±1.5 | 26.1±1.4 | 35.1±1.7 | 0.0±0.0 | Instruction-driven history-aware policies for robotic manipulations | - |
| 3D-LOTUS | 45.7 | 94.3±1.4 | 49.9±2.2 | 38.1±1.1 | 0.3±0.3 | Towards Generalizable Vision-Language Robotic Manipulation: A Benchmark and LLM-guided 3D Policy | - |
0 of 6 row(s) selected.