Robot Manipulation on LIBERO (test)
99.3Average Success RateCORALSimVLA
Evaluation results
| Method | Links | |||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| CORALSimVLABackbone=SimVLA2026.03 | 99.3 | 99.6 | 99.8 | 99 | 98.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SimVLAMode=Baseline2026.03 | 98.6 | 99.6 | 99.8 | 98.6 | 96.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| LingBot-VA2026.05 | 98.5 | 98.5 | 99.6 | 97.2 | 98.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MindPI-MoE+Lngrammodule=Lngram2026.05 | 98.5 | 99.4 | 99 | 98.8 | 96.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| LangForce2026.01 | 98.4 | 99.2 | 99.6 | 99.4 | 95.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CORALπ0.5Backbone=π0.52026.03 | 98.4 | 99.8 | 99.4 | 98.6 | 95.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| GaussianDream2026.05 | 98.4 | 99 | 99.6 | 99 | 96 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| π0.5-DiGmodule=DiG-Flow2025.12 | 98.3 | 99.2 | 99 | 98.6 | 96.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SFT2026.03 | 98.3 | 98.8 | 99 | 98.6 | 96.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MindPI-MoE2026.05 | 98.25 | 98.8 | 99 | 99 | 96.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| X-VLA2026.03 | 98.1 | 98.2 | 98.6 | 97.8 | 97.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| 3D-CAVLA2026.05 | 98.1 | 98.2 | 99.8 | 98.2 | 96.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| LaRA-VLACoT Type=Latent CoT2026.02 | 97.9 | 96.4 | 99.8 | 98.6 | 96.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PTR2026.03 | 97.8 | 98 | 99.2 | 97.6 | 97 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| InternVL3.5-ADVbackbone=InternVL3.5, mode=ADV (VLM verification into model-Diffusion), action chunk size=162026.03 | 97.8 | 97.6 | 99.4 | 97.2 | 96.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| GeoVLA2026.05 | 97.7 | 98.4 | 99 | 96.6 | 96.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SFT+Belief2026.03 | 97.6 | 98.2 | 98.4 | 98 | 95.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Spatial ForcingImplementation=PyTorch2026.05 | 97.6 | 98.6 | 98.4 | 98.2 | 95.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| AsyncVLAFine-tuning protocol=combined 4 task suites2025.11 | 97.4 | 98.4 | 99.2 | 98.6 | 93.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VLA-4D2026.05 | 97.4 | 97.9 | 98.6 | 97.8 | 94.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VLA-Adapter2026.03 | 97.3 | 97.8 | 99.2 | 97.2 | 95 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| OpenVLA-OFTtuning_type=OFT2026.01 | 97.1 | 97.6 | 98.4 | 97.9 | 94.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| OpenVLA-OFTCoT Type=No CoT2026.02 | 97.1 | 97.6 | 97.9 | 98.4 | 94.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| OpenVLA-OFT2026.01 | 97.1 | 97.6 | 98.4 | 97.9 | 94.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| OpenVLA-OFTtuning=OFT2025.12 | 97.1 | 97.6 | 98.4 | 97.9 | 94.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| OpenVLA-OFT2026.03 | 97.1 | 97.6 | 98.4 | 97.9 | 94.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| OpenVLA-OFT2026.03 | 97.1 | 97.6 | 98.4 | 97.9 | 94.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DeepThinkVLACoT Type=Textual CoT2026.02 | 97 | 99 | 96.4 | 96.6 | 96.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CogVLAGroup=Token Compression, Hardware=NVIDIA H1002026.01 | 97 | 99 | 99 | 97 | 95 | — | — | — | — | — | 2.7 | 31 | — | — | — | — | — | — | — | |
| LightVLAGroup=Token Compression, Hardware=NVIDIA H1002026.01 | 97 | 98 | 98 | 98 | 95 | — | — | — | — | — | 2.91 | 22 | — | — | — | — | — | — | — | |
| π0.5-L9Group=Layer Distillation (ours), Hardware=NVIDIA H100, Layers=9, Base Model=π0.52026.01 | 97 | 99 | 98 | 97 | 93 | — | — | — | — | — | 1.82 | 14.8 | — | — | — | — | — | — | — | |
| ELAN4DBase Model=π0.52026.05 | 97 | 98.2 | 98.8 | 96.8 | 94.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| pi0.52026.01 | 96.9 | 98.8 | 98.2 | 98 | 92.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Action-SketcherVisual Prompts=true2026.01 | 96.9 | 97.2 | 99.6 | 94.8 | 96 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| π0.52025.12 | 96.9 | 98.8 | 98.2 | 98 | 92.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| π0.5Mode=Baseline2026.03 | 96.9 | 98.8 | 98.2 | 98 | 92.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| π0.5Fine-tuning protocol=combined 4 task suites2025.11 | 96.9 | 98.8 | 98.2 | 98 | 92.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| π0.5Evaluation Protocol=fine-tuned2026.05 | 96.9 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| π0.52026.05 | 96.9 | 98.8 | 98.2 | 98 | 92.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| π0.5CoT Type=Textual CoT2026.02 | 96.8 | 98.8 | 98 | 98.2 | 92.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| π0.52026.01 | 96.8 | 98.8 | 98.2 | 98 | 92.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Qwen2.5-VL-ADVbackbone=Qwen2.5-VL, mode=ADV (VLM verification into model-Diffusion), action chunk size=162026.03 | 96.8 | 96.6 | 98.8 | 97 | 94.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MantisVision Augmentation=true2025.11 | 96.7 | 98.8 | 99.2 | 94.4 | 94.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MemoryVLA2026.03 | 96.7 | 98.4 | 98.4 | 96.4 | 93.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| π0.5-KGEvaluation Protocol=fine-tuned2026.05 | 96.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| π0.52026.05 | 96.7 | 97.8 | 98.8 | 97.6 | 92.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Qwen3-VL-OFTvariant=OFT2026.01 | 96.6 | 97.8 | 98.6 | 96.2 | 93.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| GeoPredict2026.05 | 96.6 | 98 | 98.2 | 95.7 | 94 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Qwen3-VL-GR00Tvariant=GR00T2026.01 | 96.5 | 97.8 | 98.8 | 97.4 | 92 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| GeoPredict2026.05 | 96.5 | 98 | 98.2 | 95.7 | 94 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| dVLAFine-tuning protocol=combined 4 task suites2025.11 | 96.4 | 97.4 | 97.9 | 98.2 | 92.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DD-VLA2026.03 | 96.3 | 97.2 | 98.6 | 97.4 | 92 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Discrete Diffusion VLA2026.03 | 96.3 | 97.2 | 98.6 | 97.4 | 92 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Discrete-Diffusion VLAFine-tuning protocol=combined 4 task suites2025.11 | 96.3 | 97.2 | 98.6 | 97.4 | 92 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Pri4R2026.05 | 96.3 | 93.2 | 98.6 | 98.1 | 95.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| π0.5Group=Baseline, Hardware=NVIDIA H1002026.01 | 96 | 98 | 96 | 97 | 93 | — | — | — | — | — | 3.39 | 25.5 | — | — | — | — | — | — | — | |
| EVOLVE-VLABase Model=OpenVLA-OFT, Reward Model=VLAC2025.12 | 95.8 | 95.4 | 97.4 | 95.8 | 94.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| LASTAction representation=Continuous2026.05 | 95.8 | 98.4 | 97 | 96.7 | 91.2 | — | — | — | — | — | — | — | — | — | 1 | 5 | 1 | 1 | 1 | |
| F1CoT Type=Visual CoT2026.02 | 95.7 | 98.2 | 95.4 | 97.8 | 91.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| F1Vision Augmentation=true2025.11 | 95.7 | 98.2 | 97.8 | 95.4 | 91.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| FLOWER2026.03 | 95.7 | 97.1 | 96.7 | 95.6 | 93.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| UnifiedVLAVision Augmentation=true2025.11 | 95.5 | 95.4 | 98.8 | 93.6 | 94 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| UniVLA2026.03 | 95.5 | 95.4 | 98.8 | 93.6 | 94 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| OpenVLA-OFT (Dis.)Action space=discrete, Fine-tuning protocol=combined 4 task suites2025.11 | 95.5 | 96.2 | 98.2 | 95.6 | 92 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Qwen3-VL-FASTvariant=FAST2026.01 | 95.4 | 97.3 | 97.4 | 96.3 | 90.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| OpenVLA-OFT (Con.)Action space=continuous, Fine-tuning protocol=combined 4 task suites2025.11 | 95.4 | 96.9 | 98.1 | 95.5 | 91.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| INFUSEModel Architecture=OpenVLA-7B2026.01 | 95.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| GR00T-N1-DiGmodule=DiG-Flow2025.12 | 95.3 | 96 | 98.4 | 94.8 | 92.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| OpenVLA-OFTEvaluation Protocol=fine-tuned2026.05 | 95.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| UniVLA2025.12 | 95.2 | 96.5 | 96.8 | 95.6 | 92 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| UniVLAVision Augmentation=true2025.11 | 95.2 | 96.5 | 96.8 | 95.6 | 92 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| UniVLA2026.03 | 95.2 | 96.5 | 96.8 | 95.6 | 92 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| π0Group=Baseline, Hardware=NVIDIA H1002026.01 | 95 | 97 | 97 | 93 | 92 | — | — | — | — | — | 2.93 | 22.6 | — | — | — | — | — | — | — | |
| π0-L9Group=Layer Distillation (ours), Hardware=NVIDIA H100, Layers=9, Base Model=π02026.01 | 95 | 98 | 98 | 95 | 87 | — | — | — | — | — | 1.62 | 13.5 | — | — | — | — | — | — | — | |
| π0.5-L6Group=Layer Distillation (ours), Hardware=NVIDIA H100, Layers=6, Base Model=π0.52026.01 | 95 | 98 | 96 | 94 | 90 | — | — | — | — | — | 1.3 | 11.3 | — | — | — | — | — | — | — | |
| ELAN4DBase Model=π02026.05 | 95 | 96.4 | 98.2 | 93.4 | 91.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| QDepth-VLA2026.05 | 94.9 | 97.6 | 96.6 | 95.2 | 90 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| OpenVLA-FreqPolicyNFE=12025.06 | 94.8 | 97 | 98.6 | 96 | 87.6 | — | — | — | — | — | — | — | 6.05 | — | — | — | — | — | — | |
| PD-VLA2026.03 | 94.7 | 95.5 | 96.7 | 94.9 | 91.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| InternVL3.5-Diffusionbackbone=InternVL3.5, mode=Diffusion action expert2026.03 | 94.7 | 95.8 | 96 | 94.2 | 92.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| π0-KGEvaluation Protocol=fine-tuned2026.05 | 94.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PALM2026.01 | 94.5 | 95.2 | 96.7 | 94.3 | 91.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| π02025.12 | 94.4 | 98 | 96.8 | 94.4 | 88.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| π0.5-KI2026.03 | 94.3 | 98 | 97.8 | 95.6 | 85.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| π0CoT Type=No CoT2026.02 | 94.2 | 96.8 | 95.8 | 98.8 | 85.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| pi_02025.12 | 94.2 | 96.8 | 98.8 | 95.8 | 85.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| π0Vision Augmentation=false2025.11 | 94.2 | 96.8 | 98.8 | 95.8 | 85.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| π02026.03 | 94.2 | 96.8 | 98.8 | 95.8 | 85.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| π02026.03 | 94.2 | 96.8 | 98.8 | 95.8 | 85.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| π0Fine-tuning protocol=combined 4 task suites2025.11 | 94.2 | 96.8 | 98.8 | 95.8 | 85.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| π0Evaluation Protocol=fine-tuned2026.05 | 94.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| π02026.05 | 94.2 | 96.8 | 98.8 | 95.8 | 85.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| pi02026.01 | 94.1 | 96.8 | 98.8 | 95.8 | 85.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| π02026.01 | 94.1 | 96.8 | 98.8 | 95.8 | 85.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| GLaD2025.12 | 94.1 | 95 | 97.4 | 94.4 | 89.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| π02026.05 | 94.1 | 96.8 | 98.8 | 95.8 | 85.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| π02026.05 | 94.1 | 96.8 | 98.8 | 95.8 | 85.2 | — | — | — | — | — | — | — | — | — | 2 | 1 | 2 | 6 | 2 | |
| π0-FASTAction representation=Continuous2026.05 | 94.1 | 96.6 | 97.2 | 93.1 | 89.4 | — | — | — | — | — | — | — | — | — | 3 | 4 | 5 | 2 | 2 | |
| π0-L6Group=Layer Distillation (ours), Hardware=NVIDIA H100, Layers=6, Base Model=π02026.01 | 94 | 98 | 95 | 93 | 85 | — | — | — | — | — | 1.18 | 10.5 | — | — | — | — | — | — | — | |
| GR00T-N12025.12 | 93.9 | 94.4 | 97.6 | 93 | 90.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — |