Robot Manipulation on LIBERO (Success Rates and Horizon Length)
100Spatial Success RateGeoAlign
Evaluation results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| GeoAlign2026.06 | 100 | 99.5 | 100 | 96.6 | 99 | — | — | — | — | |
| PhysBrain 1.02026.05 | 99.6 | 99.6 | 99.4 | 96.4 | 98.8 | — | — | — | — | |
| OpenVLA-OFT + VGGT-Ω Frozen Scene TokensScene Tokens=Frozen2026.05 | 99.3 | 99.2 | 99 | 96.7 | 98.5 | — | — | — | — | |
| π0.5-Async-15kinference_mode=Asynchronous, training_budget=15k steps, historical_action_conditioning=true, asynchronous_semantic_action_decoupling=true2026.06 | 99.2 | 98 | 99 | 96 | 98.1 | — | — | — | — | |
| AutoHorizonBackbone=π-0.52026.05 | 99.1 | 99.2 | 97.5 | 91.6 | 96.9 | — | — | — | — | |
| π0.52026.05 | 99 | 98 | 98 | — | 96.5 | — | 91 | — | — | |
| DA-ABL2026.05 | 99 | 98.5 | 99 | — | 97.8 | — | 94.5 | — | — | |
| GEM-VLA2026.05 | 99 | 98.8 | 97.1 | 89.3 | 96.1 | — | — | — | — | |
| VLA-ADPParadigm=VLA2026.06 | 99 | 98.2 | 96.8 | 91.2 | 96.3 | — | — | — | — | |
| GeoSem-WAMParadigm=WAM2026.06 | 99 | 100 | 98.2 | 97 | 98.55 | — | — | — | — | |
| π0.52026.05 | 98.8 | 98.2 | 98 | 92.4 | 96.9 | — | — | — | — | |
| Xiaomi-Robotics-02026.05 | 98.8 | 100 | 98.8 | 97.2 | 98.7 | — | — | — | — | |
| π0.5Paradigm=VLA2026.06 | 98.8 | 98.2 | 98 | 92.4 | 96.85 | — | — | — | — | |
| π0.5Mode=Fine-tuned, Steps=80K2026.06 | 98.8 | 98.2 | 98 | 92.4 | 96.9 | — | — | — | — | |
| π0.52026.06 | 98.8 | 98.2 | 98 | 92.4 | 96.9 | — | — | — | — | |
| π0.5inference_mode=Synchronous, training_budget=Full, historical_action_conditioning=true2026.06 | 98.8 | 98.2 | 98 | 92.4 | 96.9 | — | — | — | — | |
| π0.52026.06 | 98.8 | 98.2 | 98 | — | 96.9 | — | 92.4 | — | — | |
| OriginalBackbone=π-0.52026.05 | 98.6 | 99.8 | 97.6 | 95.4 | 97.9 | 6.3 | — | — | — | |
| A3Backbone=π-0.52026.05 | 98.6 | 99.4 | 98 | 96.2 | 98.1 | 9.7 | — | — | — | |
| SparseVLMModel=OpenVLA-OFT2026.05 | 98.6 | 91.8 | 94.8 | 92 | 94.3 | — | — | 2.289 | 52.15 | |
| DA-Vanilla2026.05 | 98.5 | 98.5 | 97 | — | 96.8 | — | 93 | — | — | |
| LingBot-VAParadigm=WAM2026.06 | 98.5 | 99.6 | 97.2 | 98.5 | 98.5 | — | — | — | — | |
| LingBot-VA*N^v=20, N^a=50, Speedup=1.0×2026.06 | 98.5 | 99.8 | 98 | 98.3 | 98.6 | — | — | — | — | |
| MoHBackbone=π-0.52026.05 | 98.4 | 99 | 98.2 | 95.2 | 97.7 | 5 | — | — | — | |
| VanillaModel=OpenVLA-OFT2026.05 | 98.4 | 98.2 | 96.4 | 94.2 | 97 | — | — | 3.97 | 70.25 | |
| APT2026.06 | 98.4 | 99.4 | 96.4 | 90.2 | 96.1 | — | — | — | — | |
| MemoryVLAMemory-augmented=true2026.06 | 98.4 | 98.4 | 96.4 | — | 96.5 | — | 93.4 | — | — | |
| A3Backbone=π-02026.05 | 98.2 | 99.2 | 96.1 | 87.2 | 95.3 | 9.4 | — | — | — | |
| EverydayVLABackbone=π-0.52026.05 | 98.2 | 99.2 | 97.4 | 95.6 | 97.6 | 6.8 | — | — | — | |
| F12026.05 | 98.2 | 97.8 | 95.4 | 91.3 | 95.7 | — | — | — | — | |
| VanillaModel=π0.52026.05 | 98.2 | 98 | 98 | 91.8 | 96.5 | — | — | 2.115 | 35.28 | |
| X-VLAN^v=–, N^a=–, Speedup=–2026.06 | 98.2 | 98.6 | 97.8 | 97.6 | 98.1 | — | — | — | — | |
| GR00T N1.5 + EQUIACTORControl Mode=Relative2026.06 | 98.2 | 95.2 | 88 | — | 91 | — | 82.6 | — | — | |
| GPC-RANK2026.05 | 98 | 99 | 97.5 | — | 96.6 | — | 92 | — | — | |
| InternVLA-M12026.05 | 98 | 99 | 93.8 | 92.6 | 95.9 | — | — | — | — | |
| π02026.05 | 98 | 96.8 | 94.4 | 88.4 | 94.4 | — | — | — | — | |
| OursModel=OpenVLA-OFT2026.05 | 98 | 98 | 96.2 | 93.4 | 96.4 | — | — | 1.722 | 37.22 | |
| Afford-VLA2026.05 | 97.8 | 99.6 | 97.6 | 94.6 | 97.4 | — | — | — | — | |
| FastVModel=OpenVLA-OFT2026.05 | 97.8 | 92.8 | 95.6 | 91.8 | 94.5 | — | — | 2.141 | 50.24 | |
| MolmoAct22026.06 | 97.8 | 100 | 97.8 | — | 97.2 | — | 93.2 | — | — | |
| GR00T N1.62026.05 | 97.7 | 98.5 | 97.5 | 94.4 | 97 | — | — | — | — | |
| GR00T-N1.72026.06 | 97.7 | 97.5 | 98.5 | — | 97 | — | 94.4 | — | — | |
| MoHBackbone=π-02026.05 | 97.6 | 98.8 | 96.4 | 87.4 | 95.1 | 5 | — | — | — | |
| OpenVLA-OFT2026.05 | 97.6 | 98.4 | 97.9 | 94.5 | 97.1 | — | — | — | — | |
| OpenVLA-OFT (L1)Action Representation=Continuous2025.08 | 97.6 | 98.4 | 97.9 | 94.5 | 97.1 | — | — | — | — | |
| OpenVLA-OFTParadigm=VLA2026.06 | 97.6 | 98.4 | 97.9 | 94.5 | 97.1 | — | — | — | — | |
| TBD-VLAMode=Fine-tuned, Steps=80K2026.06 | 97.6 | 99.6 | 97.4 | 96.6 | 97.7 | — | — | — | — | |
| OpenVLA-OFT2026.06 | 97.6 | 98.4 | 97.9 | — | 97.1 | — | 94.5 | — | — | |
| DreamVLA2026.05 | 97.5 | 94 | 89.5 | 59.5 | 92.6 | — | — | — | — | |
| DreamVLA2026.06 | 97.5 | 94 | 89.5 | 89.5 | 92.6 | — | — | — | — | |
| VLA-PrunerModel=OpenVLA-OFT2026.05 | 97.4 | 93 | 94 | 92 | 94.1 | — | — | 2.234 | 54.53 | |
| dVLAMode=Fine-tuned, Steps=80K2026.06 | 97.4 | 97.9 | 98.2 | 92.2 | 96.4 | — | — | — | — | |
| ZR-0Parameters=2.6B, Action chunk length (H)=102026.06 | 97.4 | 99.4 | 98 | — | 97.8 | — | 96.4 | — | — | |
| NORA-1.52026.06 | 97.3 | 96.4 | 94.5 | — | 94.5 | — | 89.6 | — | — | |
| OriginalBackbone=π-02026.05 | 97.2 | 98.2 | 95.6 | 84.6 | 95.1 | 5 | — | — | — | |
| π0-EqMAction Decoder=π0-EqM, Compute Budget=300-step2026.05 | 97.2 | 98.4 | 94.8 | — | 94.35 | — | 87 | — | — | |
| Qwen3VL-SFT-VLA2026.05 | 97.2 | 98.4 | 95.6 | 88.4 | 94.9 | — | — | — | — | |
| DivPruneModel=OpenVLA-OFT2026.05 | 97.2 | 90 | 93.4 | 89.6 | 92.6 | — | — | 2.126 | 53.87 | |
| Discrete Diffusion VLAAction Representation=Discretized2025.08 | 97.2 | 99.4 | 96.8 | 92.2 | 96.4 | — | — | — | — | |
| CogACTParadigm=VLA2026.06 | 97.2 | 98 | 90.2 | 88.8 | 93.55 | — | — | — | — | |
| Fast-WAMParadigm=WAM2026.06 | 97.2 | 100 | 97 | 95.2 | 97.6 | — | — | — | — | |
| Disc Diff VLAMode=Fine-tuned, Steps=80K2026.06 | 97.2 | 98.6 | 97.4 | 92 | 96.3 | — | — | — | — | |
| CogACT2026.06 | 97.2 | 98 | 90.2 | — | 93.2 | — | 88.8 | — | — | |
| π0.5-15kinference_mode=Synchronous, training_budget=15k steps, historical_action_conditioning=true2026.06 | 97.2 | 98.4 | 97.2 | 91.6 | 96.1 | — | — | — | — | |
| VLA-CacheModel=OpenVLA-OFT2026.05 | 97 | 92.2 | 94.6 | 91 | 93.7 | — | — | 2.73 | 58.14 | |
| OursModel=π0.52026.05 | 97 | 98.8 | 96 | 90.2 | 95.5 | — | — | 1.482 | 24.33 | |
| Flash-WAMN^v=1, N^a=2, Speedup=13.7×2026.06 | 97 | 92.8 | 96.4 | 98 | 95.7 | — | — | — | — | |
| VLA-0Mode=Fine-tuned, Steps=80K2026.06 | 97 | 97.8 | 96.2 | 87.6 | 94.7 | — | — | — | — | |
| OpenVLA-OFT (Diffusion)Action Representation=Continuous2025.08 | 96.9 | 98.1 | 96.2 | 91.1 | 95.6 | — | — | — | — | |
| π02026.05 | 96.8 | 98.8 | 95.8 | 85.2 | 94.2 | — | — | — | — | |
| π0Action Decoder=π0, Compute Budget=300-step2026.05 | 96.8 | 98.8 | 95.8 | — | 94.15 | — | 85.2 | — | — | |
| π02026.05 | 96.8 | 98.8 | 95.8 | 85.2 | 94.2 | — | — | — | — | |
| π0status=reported2026.05 | 96.8 | 98.8 | 95.8 | 85.2 | 94.2 | — | — | — | — | |
| π0Action Representation=Continuous2025.08 | 96.8 | 98.8 | 95.8 | 85.2 | 94.2 | — | — | — | — | |
| π0Paradigm=VLA2026.06 | 96.8 | 98.8 | 95.8 | 85.2 | 94.15 | — | — | — | — | |
| MotusParadigm=WAM2026.06 | 96.8 | 99.8 | 96.6 | 97.6 | 97.7 | — | — | — | — | |
| pi_02026.06 | 96.8 | 98.8 | 95.8 | 85.2 | 94.2 | — | — | — | — | |
| π0N^v=–, N^a=–, Speedup=–2026.06 | 96.8 | 98.8 | 95.8 | 85.2 | 94.1 | — | — | — | — | |
| π02026.06 | 96.8 | 98.8 | 95.8 | 85.2 | 94.2 | — | — | — | — | |
| π02026.06 | 96.8 | 98.8 | 95.8 | — | 94.2 | — | 85.2 | — | — | |
| π0inference_mode=Synchronous, training_budget=Full2026.06 | 96.8 | 98.8 | 95.8 | 85.2 | 94.2 | — | — | — | — | |
| π02026.06 | 96.8 | 98.8 | 95.8 | — | 94.2 | — | 85.2 | — | — | |
| GR00T 1.5 BaselineBackbone=GR00T 1.52026.06 | 96.6 | 97 | 95.4 | — | 94.9 | — | 90.6 | — | — | |
| G3VLA (π3X)Backbone=GR00T 1.5, Supervision Source=π3X2026.06 | 96.6 | 99 | 95.8 | — | 95.25 | — | 89.6 | — | — | |
| DeepThinkVLA2026.06 | 96.6 | 99 | 96.4 | — | 97 | — | 96.2 | — | — | |
| UniVLA2026.05 | 96.5 | 96.8 | 95.6 | 92 | 95.2 | — | — | — | — | |
| GR00T-N1.52026.05 | 96.5 | 98.5 | 91 | 91.5 | 94.4 | — | — | — | — | |
| UniVLAParadigm=VLA2026.06 | 96.5 | 96.8 | 95.6 | 92 | 95.23 | — | — | — | — | |
| UniVLA2026.06 | 96.5 | 96.8 | 95.6 | 92 | 95.2 | — | — | — | — | |
| UniVLAinference_mode=Synchronous, training_budget=Full2026.06 | 96.5 | 96.8 | 95.6 | 92 | 95.2 | — | — | — | — | |
| AutoHorizon2026.02 | 96.5 | 98 | 94.4 | — | — | — | 92.1 | — | — | |
| π0-FAST2026.05 | 96.4 | 96.8 | 88.6 | 60.2 | 85.5 | — | — | — | — | |
| π0-FAST2026.05 | 96.4 | 96.8 | 88.6 | 60.2 | 85.5 | — | — | — | — | |
| DepthVLA2026.05 | 96.4 | 98 | 95.8 | 89.2 | 94.9 | — | — | — | — | |
| DepthVLA2026.05 | 96.4 | 98 | 95.8 | 89.2 | 94.9 | — | — | — | — | |
| FastVModel=π0.52026.05 | 96.4 | 96.8 | 95.6 | 89 | 94.5 | — | — | 1.612 | 28.64 | |
| π0 + FASTAction Representation=Discretized2025.08 | 96.4 | 96.8 | 88.6 | 60.2 | 85.5 | — | — | — | — | |
| π0 + FASTParadigm=VLA2026.06 | 96.4 | 96.8 | 88.6 | 60.2 | 85.5 | — | — | — | — | |
| DepthVLA2026.06 | 96.4 | 98 | 95.8 | 89.2 | 94.9 | — | — | — | — | |
| π0-FastMode=Fine-tuned, Steps=80K2026.06 | 96.4 | 96.8 | 88.6 | 60.2 | 85.5 | — | — | — | — |