Robot Manipulation on LIBERO LONG (test)
97.6Success RateElasticFlow
Evaluation Results
| Method | Links | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| ElasticFlowPolicy inputs=third-person image, wrist-view image, language instruction, Official reproduction=false2026.05 | 97.6 | 93 | 97 | 99 | 99 | 99 | 100 | 100 | 98 | 91 | 100 | |
| HiF-VLAPolicy inputs=third-person image, wrist-view image, language instruction2025.12 | 96.4 | 88 | 98 | 100 | 100 | 100 | 100 | 96 | 100 | 82 | 100 | |
| HiF-VLAPolicy inputs=third-person image, wrist-view image, language instruction, Official reproduction=false2026.05 | 96.4 | 88 | 98 | 100 | 100 | 100 | 100 | 96 | 100 | 82 | 100 | |
| ElasticFlowOfficial reproduction=false2026.05 | 95.6 | 95 | 97 | 98 | 99 | 99 | 99 | 93 | 100 | 81 | 95 | |
| HiF-VLAPolicy inputs=third-view image, language instruction2025.12 | 94.4 | 94 | 98 | 100 | 100 | 94 | 100 | 90 | 98 | 76 | 94 | |
| HiF-VLAOfficial reproduction=false2026.05 | 94.4 | 94 | 98 | 100 | 100 | 94 | 100 | 90 | 98 | 76 | 94 | |
| OpenVLA-OFT*Policy inputs=third-person image, wrist-view image, language instruction2025.12 | 94 | 90 | 98 | 98 | 98 | 96 | 100 | 92 | 100 | 72 | 96 | |
| OpenVLA-OFTPolicy inputs=third-person image, wrist-view image, language instruction, Official reproduction=true2026.05 | 94 | 90 | 98 | 98 | 98 | 96 | 100 | 92 | 100 | 72 | 96 | |
| MemoryVLAPolicy inputs=third-view image, language instruction2025.12 | 93.4 | 92 | 96 | 96 | 100 | 100 | 100 | 96 | 96 | 62 | 96 | |
| MemoryVLAOfficial reproduction=false2026.05 | 93.4 | 92 | 96 | 96 | 100 | 100 | 100 | 96 | 96 | 62 | 96 | |
| OpenVLA-OFT*Policy inputs=third-view image, language instruction2025.12 | 91 | 82 | 96 | 96 | 94 | 90 | 96 | 92 | 100 | 70 | 94 | |
| OpenVLA-OFTOfficial reproduction=true2026.05 | 91 | 82 | 96 | 96 | 94 | 90 | 96 | 92 | 100 | 70 | 94 | |
| UniVLA*Policy inputs=third-person image, wrist-view image, language instruction2025.12 | 90 | 100 | 92 | 94 | 98 | 86 | 100 | 80 | 100 | 70 | 82 | |
| UniVLAPolicy inputs=third-person image, wrist-view image, language instruction, Official reproduction=true2026.05 | 90 | 100 | 92 | 94 | 98 | 86 | 100 | 80 | 100 | 70 | 82 | |
| SeerPolicy inputs=third-person image, wrist-view image, language instruction2025.12 | 87.7 | 91.7 | 90 | 98.3 | 100 | 91.7 | 93.3 | 85 | 88.3 | 61.7 | 71.7 | |
| SeerPolicy inputs=third-person image, wrist-view image, language instruction, Official reproduction=false2026.05 | 87.7 | 91.7 | 90 | 98.3 | 100 | 91.7 | 93.3 | 85 | 88.3 | 61.7 | 71.7 | |
| Seer (scratch)Policy inputs=third-person image, wrist-view image, language instruction2025.12 | 78.7 | 80 | 90 | 91.7 | 81.7 | 85 | 65 | 86.7 | 88.3 | 51.7 | 66.7 | |
| Seer (scratch)Policy inputs=third-person image, wrist-view image, language instruction, Official reproduction=false2026.05 | 78.7 | 80 | 90 | 91.7 | 81.7 | 85 | 65 | 86.7 | 88.3 | 51.7 | 66.7 | |
| DitaPre-training dataset=Droid2025.03 | 77.9 | — | — | — | — | — | — | — | — | — | — | |
| DitaPre-training dataset=OXE2025.03 | 63.8 | — | — | — | — | — | — | — | — | — | — | |
| UniVLA*Policy inputs=third-view image, language instruction2025.12 | 63 | 64 | 82 | 76 | 96 | 58 | 98 | 24 | 74 | 32 | 26 | |
| UniVLAOfficial reproduction=true2026.05 | 63 | 64 | 82 | 76 | 96 | 58 | 98 | 24 | 74 | 32 | 26 | |
| OpenVLAPolicy inputs=third-view image, language instruction2025.12 | 54 | 35 | 95 | 65 | 45 | 40 | 80 | 60 | 45 | 20 | 55 | |
| OpenVLAOfficial reproduction=false2026.05 | 54 | 35 | 95 | 65 | 45 | 40 | 80 | 60 | 45 | 20 | 55 | |
| OpenVLA2025.03 | 53.7 | — | — | — | — | — | — | — | — | — | — | |
| Octo2025.03 | 51.1 | — | — | — | — | — | — | — | — | — | — | |
| Diffusion Policy2025.03 | 50.5 | — | — | — | — | — | — | — | — | — | — |