Robotic Manipulation on LIBERO-Plus (test)
99.5Lighting Robustness ScoreSceneDiver
Evaluation results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| SceneDiver2026.06 | 99.5 | — | — | 90.61 | 76.82 | — | 64.56 | 15.72 | — | — | |
| Baseline2026.06 | 98.99 | — | — | 89.74 | 76.01 | — | 62.5 | 14.36 | — | — | |
| QWEN-ROBOTMANIP-ContextPre-training=Robotic Data, History-Context=Enabled2026.06 | 98.6 | — | 86.5 | 99.9 | 87.5 | 91.4 | 89.9 | 83.9 | 97.9 | — | |
| QuoVLAEvaluation Protocol=Zero-Shot Transfer2026.05 | 98.2 | — | 88.2 | 95.9 | 89.3 | 90.3 | 82.3 | 87.6 | 90.8 | — | |
| Libra-VLAAction Space=Hybrid, Evaluation Protocol=Zero-Shot Transfer2026.04 | 97.9 | — | 92.7 | 93.4 | 77.5 | 79.5 | 68.9 | 48.8 | 86.3 | — | |
| Being-H0.72026.06 | 97.8 | — | 82.8 | 90 | 88.5 | 84.8 | 82 | 59 | 93.5 | — | |
| ACoT-VLAEvaluation Protocol=Zero-Shot Transfer2026.05 | 97.7 | — | 87.5 | 96.5 | 88.1 | 86.6 | 72.6 | 82.6 | 87.8 | — | |
| π*0.5Action Space=Continuous, Evaluation Protocol=Supervised Fine-Tuning2026.04 | 97.3 | — | 81.1 | 94.6 | 84.9 | 75.7 | 70.3 | 41.7 | 71.8 | — | |
| π0.5Evaluation Protocol=Supervised Fine-Tuning2026.05 | 97.3 | — | 81.1 | 94.6 | 84.9 | 75.7 | 70.3 | 41.7 | 71.8 | — | |
| QWEN-ROBOTMANIPPre-training=Robotic Data2026.06 | 96.6 | — | 85.6 | 97.7 | 87.3 | 89 | 87.2 | 75.5 | 97.7 | — | |
| Cosmos-Policy2026.06 | 96.5 | — | 81.7 | 88.9 | 82.2 | 82.2 | 75.8 | 63.3 | 92.7 | — | |
| QuoVLAEvaluation Protocol=Supervised Fine-Tuning2026.05 | 96.3 | — | 83.7 | 99 | 90.4 | 90.9 | 98.4 | 73.5 | 94.8 | — | |
| ABot-M0Evaluation Protocol=Zero-Shot Transfer2026.05 | 96.2 | — | 86.4 | 91.6 | 82.6 | 80.5 | 60.4 | 67.9 | 86.4 | — | |
| π0.52026.06 | 96.2 | — | 80.8 | 94.1 | 84.5 | 84.4 | 78.4 | 73.6 | 89 | — | |
| Abot-M02026.06 | 96.2 | — | 86.4 | 91.6 | 82.6 | 80.5 | 60.4 | 67.9 | 86.4 | — | |
| QWEN-ROBOTMANIP-scratchPre-training=None2026.06 | 95.8 | — | 88.1 | 95.5 | 79.1 | 78.3 | 70.4 | 44.9 | 84.4 | — | |
| StarVLA2026.06 | 95.7 | — | 88.5 | 95.7 | 76.9 | 74.1 | 52.5 | 49.8 | 73 | — | |
| VLA-JEPAEvaluation Protocol=Zero-Shot Transfer2026.05 | 95.6 | — | 85.4 | 93.6 | 85.1 | 79.5 | 63.3 | 67.1 | 66.3 | — | |
| π0.5Evaluation Protocol=Zero-Shot Transfer2026.05 | 95.5 | — | 83.3 | 95 | 87 | 85.7 | 75.8 | 79.4 | 89.6 | — | |
| Libra-VLAAction Space=Hybrid, Evaluation Protocol=Supervised Fine-Tuning2026.04 | 95.3 | — | 83.2 | 94.3 | 75.3 | 82.3 | 94.5 | 41.8 | 93.7 | — | |
| ACoT-VLAEvaluation Protocol=Supervised Fine-Tuning2026.05 | 95.1 | — | 79.7 | 97.1 | 85 | 88 | 96.6 | 70.4 | 95.9 | — | |
| OpenVLA-OFT+Action Space=Continuous, Evaluation Protocol=Supervised Fine-Tuning2026.04 | 94.9 | — | 85.8 | 93.9 | 77.6 | 79.6 | 92.8 | 30.3 | 89.3 | — | |
| HiMem-WAMSetting=Zero-Shot, Training Data=standard LIBERO dataset, Stage II Pretraining=true2026.06 | 92.2 | 38.1 | 76.6 | 91 | 74.9 | 76 | 78.2 | — | 80.7 | — | |
| HiMem-WAM (w/o Stage II)Setting=Zero-Shot, Training Data=standard LIBERO dataset, Stage II Pretraining=false2026.06 | 91.1 | 37.9 | 71.7 | 83.6 | 70.7 | 72.2 | 77.2 | — | 73.2 | — | |
| PionOptimizer=Pion2026.05 | 90.67 | — | 86.93 | 84.53 | 76.71 | — | 70.88 | 63.18 | 76.09 | 75.93 | |
| π0Evaluation Protocol=Zero-Shot Transfer2026.05 | 89.3 | — | 63.5 | 84.1 | 76.4 | 69.4 | 61 | 40.8 | 80.1 | — | |
| OpenVLA-OFTAction Space=Continuous, Evaluation Protocol=Zero-Shot Transfer2026.04 | 88.7 | — | 79.5 | 93.3 | 74.2 | 69.6 | 56.4 | 31.9 | 75.8 | — | |
| OpenVLA-OFTEvaluation Protocol=Zero-Shot Transfer2026.05 | 88.7 | — | 79.5 | 93.3 | 74.2 | 69.6 | 56.4 | 31.9 | 75.8 | — | |
| OpenVLA-OFTSetting=Zero-Shot, Training Data=standard LIBERO dataset, Method Category=Vision-Language-Action Models2026.06 | 88.7 | 31.9 | 79.5 | 93.3 | 74.2 | 71.4 | 56.4 | — | 75.8 | — | |
| RIPT-VLASetting=Zero-Shot, Training Data=standard LIBERO dataset, Method Category=Vision-Language-Action Models2026.06 | 88.4 | 31.2 | 77.6 | 91.6 | 74.2 | 70.2 | 55.2 | — | 73.5 | — | |
| HoloBrain-0Setting=Zero-Shot, Training Data=standard LIBERO dataset, Method Category=World Action Models2026.06 | 88.1 | 58.2 | 78.7 | 90.3 | 79.5 | 75.3 | 65.5 | — | 66.9 | — | |
| MuonOptimizer=Muon2026.05 | 86.17 | — | 77.53 | 82.72 | 76.21 | — | 68 | 57.36 | 69.98 | 72.34 | |
| π0Setting=Zero-Shot, Training Data=standard LIBERO dataset, Method Category=Vision-Language-Action Models2026.06 | 85 | 6 | 58.8 | 81.4 | 68.9 | 56.1 | 13.8 | — | 79 | — | |
| π02026.06 | 85 | — | 58.8 | 81.4 | 68.9 | 53.6 | 13.8 | 6 | 79 | — | |
| π*0Action Space=Continuous, Evaluation Protocol=Supervised Fine-Tuning2026.04 | 84.7 | — | 72.5 | 86.2 | 69.4 | 67.4 | 79.6 | 21.1 | 68.3 | — | |
| π0Evaluation Protocol=Supervised Fine-Tuning2026.05 | 84.7 | — | 72.5 | 86.2 | 69.4 | 67.4 | 79.6 | 21.1 | 68.3 | — | |
| AdamWOptimizer=AdamW2026.05 | 76.35 | — | 54.5 | 68.97 | 61.8 | — | 70.38 | 47.04 | 66.37 | 64.57 | |
| π0-FastAction Space=Discrete, Evaluation Protocol=Zero-Shot Transfer2026.04 | 73.2 | — | 61 | 73.2 | 68.8 | 61.6 | 65.1 | 21.6 | 74.4 | — | |
| UniVLAAction Space=Continuous, Evaluation Protocol=Zero-Shot Transfer2026.04 | 69 | — | 69.6 | 81 | 31.9 | 42.9 | 1.8 | 46.2 | 21.2 | — | |
| UniVLAEvaluation Protocol=Zero-Shot Transfer2026.05 | 69 | — | 69.6 | 81 | 31.9 | 42.9 | 1.8 | 46.2 | 21.2 | — | |
| UniVLASetting=Zero-Shot, Training Data=standard LIBERO dataset, Method Category=Vision-Language-Action Models2026.06 | 69 | 46.2 | 69.6 | 81 | 31.9 | 45.8 | 1.8 | — | 21.2 | — | |
| NORAAction Space=Discrete, Evaluation Protocol=Zero-Shot Transfer2026.04 | 45.7 | — | 65.1 | 58.6 | 62.1 | 39 | 2.2 | 37 | 12.8 | — | |
| WorldVLAAction Space=Discrete, Evaluation Protocol=Zero-Shot Transfer2026.04 | 43.7 | — | 41.6 | 17.1 | 38 | 25 | 0.1 | 27.9 | 10.9 | — | |
| WorldVLASetting=Zero-Shot, Training Data=standard LIBERO dataset, Method Category=World Action Models2026.06 | 43.7 | 27.9 | 41.6 | 17.1 | 38 | 25.6 | 0.1 | — | 10.9 | — | |
| SmoothVLATraining Protocol=DPO2026.03 | 17.5 | 85.2 | 40.1 | 58.6 | 62.9 | 44.8 | — | — | — | — | |
| OpenVLATraining Protocol=DPO2026.03 | 16.5 | 84.2 | 39.9 | 42.1 | 58.2 | 39.2 | — | — | — | — | |
| SmoothVLATraining Protocol=GRPO2026.03 | 16.1 | 85.6 | 43.3 | 61.2 | 64.4 | 46.2 | — | — | — | — | |
| OpenVLATraining Protocol=GRPO2026.03 | 15.4 | 84.5 | 36.7 | 48.8 | 55.7 | 39.2 | — | — | — | — | |
| OpenVLAAction Space=Discrete, Evaluation Protocol=Zero-Shot Transfer2026.04 | 8.1 | — | 23 | 34.8 | 28.5 | 15.6 | 0.8 | 3.5 | 15.2 | — | |
| OpenVLAEvaluation Protocol=Zero-Shot Transfer2026.05 | 8.1 | — | 23 | 34.8 | 28.5 | 15.6 | 0.8 | 3.5 | 15.2 | — | |
| OpenVLASetting=Zero-Shot, Training Data=standard LIBERO dataset, Method Category=Vision-Language-Action Models2026.06 | 8.1 | 3.5 | 23 | 34.8 | 28.5 | 16.3 | 0.8 | — | 15.2 | — | |
| OpenVLATraining Protocol=SFT2026.03 | 4.4 | 82.1 | 26.8 | 25.3 | 31.6 | 22 | — | — | — | — | |
| OpenVLA-OFTEvaluation Protocol=zero-shot transfer from LIBERO2026.05 | — | — | — | — | — | — | — | — | — | 69.6 | |
| OpenVLA-OFTEvaluation Protocol=fine-tuned2026.05 | — | — | — | — | — | — | — | — | — | 79.6 | |
| π0Evaluation Protocol=zero-shot transfer from LIBERO2026.05 | — | — | — | — | — | — | — | — | — | 53.6 | |
| π0Evaluation Protocol=fine-tuned2026.05 | — | — | — | — | — | — | — | — | — | 84 | |
| π0-KGEvaluation Protocol=zero-shot transfer from LIBERO2026.05 | — | — | — | — | — | — | — | — | — | 72.8 | |
| π0-KGEvaluation Protocol=fine-tuned2026.05 | — | — | — | — | — | — | — | — | — | 88.5 | |
| π0.5Evaluation Protocol=zero-shot transfer from LIBERO2026.05 | — | — | — | — | — | — | — | — | — | 83.9 | |
| π0.5Evaluation Protocol=fine-tuned2026.05 | — | — | — | — | — | — | — | — | — | 90.4 | |
| π0.5-KGEvaluation Protocol=zero-shot transfer from LIBERO2026.05 | — | — | — | — | — | — | — | — | — | 87.7 | |
| π0.5-KGEvaluation Protocol=fine-tuned2026.05 | — | — | — | — | — | — | — | — | — | 92.6 |