Robotic Manipulation on LIBERO Spatial Object Goal Long
98.1Overall Success Rate (Long)Agentic-VLA
Evaluation Results
| Method | Links | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Agentic-VLALearning Paradigm=RL Methods, Reproduced=true2026.05 | 98.1 | 97.2 | 98.6 | 97.4 | 97.8 | — | — | — | — | — | — | — | |
| LaMPMotion Guidance=Enabled2026.03 | 96.7 | 99.4 | 99.8 | 97.4 | 98.3 | — | — | — | — | — | — | — | |
| ActQuantModel=OpenVLA-OFT, Vision+LLM BPW=3.52026.05 | 95.6 | 95.8 | 98.4 | 96 | 96.5 | — | — | — | — | 0.4 | 3.8 | 3.4 | |
| ActQuantModel=OpenVLA-OFT, Vision+LLM BPW=3.02026.05 | 95.6 | 92.8 | 97.4 | 94 | 95 | — | — | — | — | 1.9 | 3.5 | 3.2 | |
| QVLAModel=OpenVLA-OFT, Vision+LLM BPW=4.02026.05 | 95.2 | 96 | 98.7 | 97.4 | 96.8 | — | — | — | — | 0.1 | 4.8 | 4.5 | |
| BaselineModel=OpenVLA-OFT, Vision+LLM BPW=16.02026.05 | 95.1 | 97.6 | 98.4 | 96.8 | 96.9 | — | — | — | — | 0 | 16 | 14.3 | |
| ActQuantModel=OpenVLA-OFT, Vision+LLM BPW=4.02026.05 | 95 | 96 | 98.2 | 97.2 | 96.6 | — | — | — | — | 0.3 | 4.4 | 4.2 | |
| VLASH2026.03 | 94.6 | 97.5 | 99.2 | 97.3 | 97.1 | 40.6 | — | — | — | — | — | — | |
| OpenVLA-OFTArchitecture Category=General VLA2026.03 | 94.5 | 97.6 | 98.4 | 97.9 | 97.1 | — | — | — | — | — | — | — | |
| EVOLVE-VLALearning Paradigm=RL Methods, Reproduced=true2026.05 | 94.4 | 95.4 | 97.4 | 95.8 | 95.8 | — | — | — | — | — | — | — | |
| AWQModel=OpenVLA-OFT, Vision+LLM BPW=4.02026.05 | 94.4 | 94.6 | 98.6 | 96.8 | 96.1 | — | — | — | — | 0.8 | 4.3 | 4.1 | |
| Sync (every step)Backbone=π0.5, VLM Backbone=46ms, Action Head (1-step)=30ms, Bridge=6ms (SigLIP 5ms + DiT 1ms), f=42026.05 | 93.83 | 98.83 | 98.17 | 97 | 96.96 | 76 | — | 31 | 1 | — | — | — | |
| Latent BridgeBackbone=π0.5, VLM Backbone=46ms, Action Head (1-step)=30ms, Bridge=6ms (SigLIP 5ms + DiT 1ms), f=42026.05 | 93.67 | 99 | 97.67 | 97.33 | 96.92 | 46 | — | 33 | 1.65 | — | — | — | |
| QVLAModel=OpenVLA-OFT, Vision+LLM BPW=3.52026.05 | 93.6 | 82.9 | 97.2 | 91.4 | 91.3 | — | — | — | — | 5.6 | 4.3 | 4.1 | |
| BaselineModel=pi0.5, Vision+LLM BPW=16.02026.05 | 93.6 | 98.4 | 98.4 | 97.6 | 97 | — | — | — | — | 0 | 16 | 6.7 | |
| StreamingVLAAFM=true2026.03 | 93 | 97.6 | 99 | 98.8 | 97.1 | 33.7 | 76.1 | — | — | — | — | — | |
| Sync (every step)Backbone=GR00T-N1.6-3B, VLM Backbone=63ms, Action Head (1-step)=27ms, Bridge=2ms, f=2–4 (phase-aware)2026.05 | 93 | 96.17 | 99.83 | 97.33 | 96.58 | 90 | — | 19 | 1 | — | — | — | |
| AWQModel=pi0.5, Vision+LLM BPW=4.02026.05 | 93 | 96.6 | 98 | 96.8 | 96.1 | — | — | — | — | 0.9 | 5.6 | 2.4 | |
| GPTQModel=OpenVLA-OFT, Vision+LLM BPW=4.02026.05 | 92.6 | 90.8 | 98.8 | 96.4 | 94.6 | — | — | — | — | 2.3 | 4.3 | 4.1 | |
| QVLAModel=pi0.5, Vision+LLM BPW=3.52026.05 | 92.6 | 97.8 | 96.2 | 91.4 | 94.5 | — | — | — | — | 2.5 | 6 | 2.5 | |
| π0.5Architecture Category=General VLA2026.03 | 92.4 | 98.8 | 98.2 | 98 | 96.9 | — | — | — | — | — | — | — | |
| π0.5h=52026.03 | 92.4 | 98.8 | 98.2 | 98 | 96.9 | 74.5 | 232.3 | — | — | — | — | — | |
| ActQuantModel=pi0.5, Vision+LLM BPW=3.52026.05 | 92.4 | 99 | 99.6 | 94 | 96.3 | — | — | — | — | 0.7 | 5.8 | 2.5 | |
| LAFMNumber of Parameters=0.11B, VLA Pre-training (VLA Pt.)=No2026.06 | 92.3 | 97.2 | 99.2 | 97.2 | 96.5 | — | — | — | — | — | — | — | |
| UniVLAArchitecture Category=Latent-Action VLA2026.03 | 92 | 96.5 | 96.8 | 95.6 | 95.2 | — | — | — | — | — | — | — | |
| UniVLANumber of Parameters=7B, VLA Pre-training (VLA Pt.)=Yes2026.06 | 92 | 96.5 | 96.8 | 95.6 | 95.2 | — | — | — | — | — | — | — | |
| GPTQModel=pi0.5, Vision+LLM BPW=4.02026.05 | 91.8 | 97.2 | 96.8 | 97 | 95.7 | — | — | — | — | 1.3 | 5.6 | 2.4 | |
| QVLAModel=pi0.5, Vision+LLM BPW=4.02026.05 | 91.8 | 98 | 97.2 | 96.4 | 95.8 | — | — | — | — | 1.2 | 6.3 | 2.7 | |
| ActQuantModel=pi0.5, Vision+LLM BPW=4.02026.05 | 91.8 | 98.4 | 99.4 | 96.8 | 96.6 | — | — | — | — | 0.4 | 6.2 | 2.7 | |
| AWQModel=OpenVLA-OFT, Vision+LLM BPW=3.02026.05 | 91.4 | 86.4 | 98 | 90.2 | 91.5 | — | — | — | — | 5.4 | 3.4 | 3.2 | |
| F1Architecture Category=Video-Based VLA2026.03 | 91.3 | 98.2 | 97.8 | 95.4 | 95.7 | — | — | — | — | — | — | — | |
| FMNumber of Parameters=0.11B, VLA Pre-training (VLA Pt.)=No2026.06 | 91.2 | 96.1 | 97.9 | 96.5 | 95.4 | — | — | — | — | — | — | — | |
| StreamingVLAAFM=true, AEO=true2026.03 | 91 | 96.6 | 96.6 | 95.4 | 94.9 | 31.625 | 36 | — | — | — | — | — | |
| ActQuantModel=OpenVLA-OFT, Vision+LLM BPW=2.52026.05 | 91 | 86.4 | 98.2 | 84.8 | 90.1 | — | — | — | — | 6.8 | 3 | 2.7 | |
| ACTNumber of Parameters=0.05B, VLA Pre-training (VLA Pt.)=No2026.06 | 90.8 | 96.3 | 99.5 | 97 | 95.9 | — | — | — | — | — | — | — | |
| GR00T N1Architecture Category=General VLA2026.03 | 90.6 | 94.4 | 97.6 | 93 | 93.9 | — | — | — | — | — | — | — | |
| AWQModel=pi0.5, Vision+LLM BPW=3.02026.05 | 90.2 | 93.4 | 97 | 91.4 | 93 | — | — | — | — | 4 | 5 | 2.1 | |
| SmolVLA2026.03 | 90 | 97 | 99 | 97 | 95.8 | 51.1 | 180.7 | — | — | — | — | — | |
| Latent BridgeBackbone=GR00T-N1.6-3B, VLM Backbone=63ms, Action Head (1-step)=27ms, Bridge=2ms, f=2–4 (phase-aware)2026.05 | 89.17 | 95.83 | 97.83 | 95.33 | 94.54 | 49 | — | 20 | 1.73 | — | — | — | |
| GPTQModel=pi0.5, Vision+LLM BPW=3.02026.05 | 89 | 92.2 | 95.4 | 88.6 | 91.3 | — | — | — | — | 5.7 | 5 | 2.1 | |
| π0.5h=102026.03 | 88.6 | 97.4 | 98.2 | 96.2 | 95.1 | 49.9 | 230.8 | — | — | — | — | — | |
| SimpleVLA-RLLearning Paradigm=RL Methods, Reproduced=true2026.05 | 87.7 | 94.3 | 90.5 | 92.3 | 91.2 | — | — | — | — | — | — | — | |
| GPTQModel=OpenVLA-OFT, Vision+LLM BPW=3.02026.05 | 87.2 | 82.4 | 98 | 88.8 | 89 | — | — | — | — | 7.9 | 3.4 | 3.2 | |
| ActQuantModel=pi0.5, Vision+LLM BPW=3.02026.05 | 87.2 | 98.2 | 98.8 | 95 | 94.8 | — | — | — | — | 2.2 | 5.6 | 2.4 | |
| VLA-CacheBackbone=π0.5, VLM Backbone=46ms, Action Head (1-step)=30ms, Bridge=6ms (SigLIP 5ms + DiT 1ms), f=42026.05 | 86.83 | 93.17 | 91 | 92.17 | 90.79 | 65 | — | 33 | 1.08 | — | — | — | |
| QVLAModel=pi0.5, Vision+LLM BPW=3.02026.05 | 86.8 | 97.4 | 97.8 | 87.8 | 92.5 | — | — | — | — | 4.5 | 5.6 | 2.4 | |
| OpenVLA-OFTLearning Paradigm=SFT Methods, Reproduced=true2026.05 | 85.8 | 91.3 | 90.1 | 89.8 | 89.2 | — | — | — | — | — | — | — | |
| π0Architecture Category=General VLA2026.03 | 85.2 | 96.8 | 98.8 | 95.8 | 94.2 | — | — | — | — | — | — | — | |
| π0Learning Paradigm=SFT Methods2026.05 | 85.2 | 96.8 | 98.8 | 95.8 | 94.2 | — | — | — | — | — | — | — | |
| VLA-CacheBackbone=GR00T-N1.6-3B, VLM Backbone=63ms, Action Head (1-step)=27ms, Bridge=2ms, f=2–4 (phase-aware)2026.05 | 85.17 | 91.17 | 90 | 89 | 88.84 | 75 | — | 21 | 1.08 | — | — | — | |
| StreamingVLAAFM=true, ANAO=true2026.03 | 84.8 | 86 | 96.2 | 93 | 90 | 30.775 | 27.75 | — | — | — | — | — | |
| Temporal Ensembling2026.03 | 78.6 | 97.6 | 92.4 | 91.4 | 90 | 279 | 231.6 | — | — | — | — | — | |
| LaMP (w/o motion)Motion Guidance=Disabled2026.03 | 78.2 | 95.8 | 98.9 | 96.6 | 92.4 | — | — | — | — | — | — | — | |
| StreamingVLAAFM=true, NEO=true2026.03 | 78.2 | 80.6 | 93.8 | 92.2 | 86.2 | 29.3 | 23 | — | — | — | — | — | |
| SmolVLANumber of Parameters=2.25B, VLA Pre-training (VLA Pt.)=No2026.06 | 77 | 93 | 94 | 91 | 88.8 | — | — | — | — | — | — | — | |
| villa-XArchitecture Category=Latent-Action VLA2026.03 | 74.5 | 97.5 | 97 | 91.5 | 90.1 | — | — | — | — | — | — | — | |
| ActQuantModel=pi0.5, Vision+LLM BPW=2.52026.05 | 73.6 | 96.6 | 98.4 | 74 | 85.7 | — | — | — | — | 11.3 | 5.2 | 2.2 | |
| π0Number of Parameters=3.3B, VLA Pre-training (VLA Pt.)=Yes2026.06 | 73 | 90 | 86 | 95 | 86 | — | — | — | — | — | — | — | |
| FlowVLAArchitecture Category=2D Flow/Trace-Guided VLA2026.03 | 72.6 | 93.2 | 95 | 91.6 | 88.1 | — | — | — | — | — | — | — | |
| SCALEBackbone=OpenVLA, Training Regime=Fine-tuned, Inference Protocol=Single inference, Decoding Strategy=SCALE2026.02 | 63.3 | 89.5 | 91 | 82.3 | 81.5 | — | — | — | — | — | — | — | |
| FastVBackbone=π0.5, VLM Backbone=46ms, Action Head (1-step)=30ms, Bridge=6ms (SigLIP 5ms + DiT 1ms), f=42026.05 | 62.17 | 88.17 | 86 | 86.83 | 80.79 | 67 | — | 37 | 0.95 | — | — | — | |
| π0 + FASTLearning Paradigm=SFT Methods2026.05 | 60.2 | 96.4 | 96.8 | 88.6 | 85.5 | — | — | — | — | — | — | — | |
| TACOBackbone=OpenVLA (fine-tuned), Inference Protocol=Test-time scaling, Training Regime=Training-required2026.02 | 60 | — | — | — | — | — | — | — | — | — | — | — | |
| WorldVLAArchitecture Category=Video-Based VLA2026.03 | 60 | 87.6 | 96.2 | 83.4 | 81.8 | — | — | — | — | — | — | — | |
| VLA-RLLearning Paradigm=RL Methods2026.05 | 59.8 | 90.2 | 91.8 | 82.2 | 81 | — | — | — | — | — | — | — | |
| FastVBackbone=GR00T-N1.6-3B, VLM Backbone=63ms, Action Head (1-step)=27ms, Bridge=2ms, f=2–4 (phase-aware)2026.05 | 58.83 | 86.17 | 84 | 85 | 78.5 | 77 | — | 23 | 0.95 | — | — | — | |
| RoboMonkeyBackbone=OpenVLA (fine-tuned), Inference Protocol=Test-time scaling, Training Regime=Training-required2026.02 | 56.5 | — | — | — | — | — | — | — | — | — | — | — | |
| MG-SelectBackbone=OpenVLA (fine-tuned), Inference Protocol=Test-time scaling, Training Regime=Training-required2026.02 | 55.4 | 81.7 | 72.5 | 73.6 | 70.8 | — | — | — | — | — | — | — | |
| OpenVLA* + Top-kBackbone=OpenVLA, Training Regime=Fine-tuned, Inference Protocol=Single inference, Decoding Strategy=Top-k, k=40, Temperature=0.72026.02 | 55.2 | 85.2 | 88.2 | 78.3 | 76.7 | — | — | — | — | — | — | — | |
| OpenVLA* + Top-pBackbone=OpenVLA, Training Regime=Fine-tuned, Inference Protocol=Single inference, Decoding Strategy=Top-p, p=0.92026.02 | 55.1 | 86.9 | 88.1 | 78.6 | 77.2 | — | — | — | — | — | — | — | |
| QVLAModel=pi0.5, Vision+LLM BPW=2.52026.05 | 55 | 93 | 97.6 | 71 | 79.2 | — | — | — | — | 17.8 | 5.3 | 2.2 | |
| OpenVLA* + SamplingBackbone=OpenVLA, Training Regime=Fine-tuned, Inference Protocol=Single inference, Decoding Strategy=Sampling, Temperature=1.02026.02 | 54.7 | 85.1 | 87.9 | 78.9 | 76.7 | — | — | — | — | — | — | — | |
| TraceVLAArchitecture Category=2D Flow/Trace-Guided VLA2026.03 | 54.1 | 84.6 | 85.2 | 75.1 | 75.8 | — | — | — | — | — | — | — | |
| OpenVLAArchitecture Category=General VLA2026.03 | 53.7 | 84.7 | 88.4 | 79.2 | 76.5 | — | — | — | — | — | — | — | |
| OpenVLALearning Paradigm=SFT Methods2026.05 | 53.7 | 84.7 | 88.4 | 79.2 | 76.5 | — | — | — | — | — | — | — | |
| OpenVLANumber of Parameters=7B, VLA Pre-training (VLA Pt.)=Yes2026.06 | 53.7 | 84.7 | 88.4 | 79.2 | 76.5 | — | — | — | — | — | — | — | |
| OpenVLA*Backbone=OpenVLA, Training Regime=Fine-tuned, Inference Protocol=Single inference, Decoding Strategy=Greedy2026.02 | 52.7 | 86.2 | 86.2 | 77.7 | 75.7 | — | — | — | — | — | — | — | |
| Feature CachingBackbone=π0.5, VLM Backbone=46ms, Action Head (1-step)=30ms, Bridge=6ms (SigLIP 5ms + DiT 1ms), f=42026.05 | 52.67 | 46.67 | 57.33 | 68.83 | 56.38 | 45 | — | — | — | — | — | — | |
| OctoLearning Paradigm=SFT Methods2026.05 | 51.1 | 78.9 | 85.7 | 84.6 | 75.1 | — | — | — | — | — | — | — | |
| OctoNumber of Parameters=0.09B, VLA Pre-training (VLA Pt.)=Yes2026.06 | 51.1 | 78.9 | 85.7 | 84.6 | 75.1 | — | — | — | — | — | — | — | |
| DPNumber of Parameters=0.26B, VLA Pre-training (VLA Pt.)=No2026.06 | 50.5 | 78.3 | 92.5 | 68.3 | 72.4 | — | — | — | — | — | — | — | |
| Feature CachingBackbone=GR00T-N1.6-3B, VLM Backbone=63ms, Action Head (1-step)=27ms, Bridge=2ms, f=2–4 (phase-aware)2026.05 | 42.33 | 35.67 | 3 | 56 | 34.25 | 48 | — | — | — | — | — | — | |
| QVLAModel=OpenVLA-OFT, Vision+LLM BPW=3.02026.05 | 31.8 | 40.8 | 56.8 | 20.2 | 37.4 | — | — | — | — | 59.5 | 3.9 | 3.7 | |
| RTCd=12026.03 | 25.2 | 96.2 | 19.8 | 93 | 58.55 | 50.2 | 203.6 | — | — | — | — | — | |
| ActQuantModel=pi0.5, Vision+LLM BPW=2.02026.05 | 25 | 61.2 | 80 | 25 | 48 | — | — | — | — | 49 | 4.8 | 2 | |
| AWQModel=OpenVLA-OFT, Vision+LLM BPW=2.02026.05 | 0 | 0 | 0 | 0 | 0 | — | — | — | — | 96.9 | 2.7 | 2.4 | |
| GPTQModel=OpenVLA-OFT, Vision+LLM BPW=2.02026.05 | 0 | 0 | 0 | 0 | 0 | — | — | — | — | 96.9 | 2.7 | 2.4 | |
| QVLAModel=OpenVLA-OFT, Vision+LLM BPW=2.52026.05 | 0 | 0 | 0.2 | 0 | 0 | — | — | — | — | 96.9 | 3.4 | 3.1 | |
| AWQModel=pi0.5, Vision+LLM BPW=2.02026.05 | 0 | 0 | 0 | 0 | 0 | — | — | — | — | 97 | 4.1 | 1.7 | |
| GPTQModel=pi0.5, Vision+LLM BPW=2.02026.05 | 0 | 0 | 0 | 0 | 0 | — | — | — | — | 97 | 4.1 | 1.7 | |
| QVLAModel=pi0.5, Vision+LLM BPW=2.02026.05 | 0 | 0 | 0 | 0 | 0 | — | — | — | — | 97 | 4.7 | 2 | |
| mimic-videoArchitecture Category=Video-Based VLA2026.03 | — | 94.2 | 96.8 | 90.6 | 93.9 | — | — | — | — | — | — | — |