Robotic Manipulation on SimplerEnv Google Robot Tasks Visual Matching and Variant Aggregation
85.7VM - Pick Coke Can Success RateRT-1 (converged)
Evaluation results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| RT-1 (converged)Model Checkpoint Stage=converged, Training Dataset=Fractal2026.04 | 85.7 | 44.2 | 73 | 74.6 | 89.8 | 50 | 32.3 | 63.3 | — | |
| Discrete Diffusion VLAPre-training=OXE, Fine-tuning=Fractal2025.08 | 85.4 | 67.5 | 60.6 | 71.2 | 82.5 | 64.6 | 23.6 | 56.9 | 64.1 | |
| ReFineVLA2026.04 | 83 | 95.3 | 51.4 | 76.6 | 83.8 | 88.1 | 34.4 | 68.8 | — | |
| SpatialVLA (fine-tuning)Evaluation Protocol=fine-tuning, Training Dataset=Fractal2026.04 | 82.5 | 85.7 | 54.6 | 74.3 | 90.6 | 79.1 | 26.2 | 65.3 | — | |
| SpatialVLA (zero-shot)Evaluation Protocol=zero-shot, Training Dataset=OXE2026.04 | 81.7 | 85.7 | 56 | 74.4 | 89.7 | 79.7 | 33 | 67.4 | — | |
| RT-2-XTraining Dataset=OXE, Evaluation Protocol=zero-shot2026.04 | 78.7 | 77.9 | 25 | 60.7 | 82.3 | 79.2 | 35.3 | 64.3 | — | |
| RT-2-XPre-training=OXE, Fine-tuning=Fractal2025.08 | 78.7 | 77.9 | 25 | 60.7 | 82.3 | 79.2 | 35.3 | 64.3 | 62.5 | |
| RoboVLM (fine-tuning)Evaluation Protocol=fine-tuning, Training Dataset=Fractal2026.04 | 77.3 | 61.7 | 43.5 | 63.4 | 75.6 | 60 | 10.6 | 51.3 | — | |
| RoboVLMPre-training=OXE, Fine-tuning=Fractal2025.08 | 77.3 | 61.7 | 43.5 | 63.4 | 75.6 | 60 | 10.6 | 51.3 | 57.4 | |
| π0-FASTPre-training=OXE, Fine-tuning=Fractal2025.08 | 75.3 | 67.5 | 42.9 | 61.9 | 77.6 | 68.2 | 31.3 | 59 | 60.5 | |
| MotoPre-training=OXE, Fine-tuning=Fractal2025.08 | 74 | 60.4 | 43.1 | 59.2 | — | — | — | — | — | |
| RoboVLM (zero-shot)Evaluation Protocol=zero-shot, Training Dataset=OXE2026.04 | 72.7 | 66.3 | 26.8 | 56.3 | 68.3 | 56 | 8.5 | 46.3 | — | |
| π0Pre-training=OXE, Fine-tuning=Fractal2025.08 | 72.7 | 65.3 | 38.3 | 58.8 | 75.2 | 63.7 | 25.6 | 54.8 | 56.8 | |
| OpenVLA-OFTPre-training=OXE, Fine-tuning=Fractal2025.08 | 72.3 | 69.6 | 47.2 | 63 | 65.3 | 59 | 12.2 | 45.5 | 54.3 | |
| RT-1 (15%)Model Checkpoint Stage=15%, Training Dataset=Fractal2026.04 | 71 | 35.4 | 56.5 | 60.2 | 81.3 | 44.6 | 26.7 | 56.2 | — | |
| RT-1-XTraining Dataset=OXE, Evaluation Protocol=zero-shot2026.04 | 56.7 | 31.7 | 59.7 | 53.4 | 49 | 32.3 | 29.4 | 39.6 | — | |
| RT-1-XPre-training=OXE, Fine-tuning=Fractal2025.08 | 56.7 | 31.7 | 59.7 | 53.4 | 49 | 32.3 | 29.4 | 39.6 | 46.5 | |
| HPT2026.04 | 56 | 60 | 24 | 46 | — | — | — | — | — | |
| HPTPre-training=OXE, Fine-tuning=Fractal2025.08 | 56 | 60 | 24 | 46 | — | — | — | — | — | |
| GR00T-N1Pre-training=OXE, Fine-tuning=Fractal2025.08 | 47 | 70 | 18.1 | 45 | 78.8 | 62.5 | 13.2 | 51.5 | 48.4 | |
| TraceVLA2026.04 | 28 | 53.7 | 57 | 42 | 60 | 56.4 | 31 | 45 | — | |
| TraceVLAPre-training=OXE, Fine-tuning=Fractal2025.08 | 28 | 53.7 | 57 | 42 | 60 | 56.4 | 31 | 45 | 43.5 | |
| Octo-BaseTraining Dataset=OXE, Evaluation Protocol=zero-shot2026.04 | 17 | 4.2 | 22.7 | 16.8 | 0.6 | 3.1 | 1.1 | 1.1 | — | |
| Octo-BasePre-training=OXE, Fine-tuning=Fractal2025.08 | 17 | 4.2 | 22.7 | 16.8 | 0.6 | 3.1 | 1.1 | 1.1 | 9 | |
| OpenVLATraining Dataset=OXE, Evaluation Protocol=zero-shot2026.04 | 16.3 | 46.2 | 35.6 | 27.7 | 54.5 | 47.7 | 17.7 | 39.8 | — | |
| OpenVLAPre-training=OXE, Fine-tuning=Fractal2025.08 | 16.3 | 46.2 | 35.6 | 27.7 | 54.5 | 47.7 | 17.7 | 39.8 | 33.8 | |
| RT-1 (begin)Model Checkpoint Stage=begin, Training Dataset=Fractal2026.04 | 2.7 | 5 | 13.9 | 6.8 | 2.2 | 4 | 6.9 | 4.2 | — |