LIBERO-Plus leaderboard
Robustness variant of LIBERO that perturbs camera, robot initial state, language, lighting, background, sensor noise and layout. Models trained on standard LIBERO are tested without further training.
Simulation · Franka Panda · MuJoCo (robosuite)Report an error
93.2%Best · InternW0-Δ
37Models ranked
OpenSaturation
Nov 18, 2026Numbers checked
| # | Model | Type | Score | Date | Source |
|---|---|---|---|---|---|
| 1 | World-action | 93.2%, self-reported by the authors, not a leaderboard entry | Sep 2026 | ||
| 2 | Agent system | 92.4%, self-reported by the authors, not a leaderboard entry | Jul 2026 | ||
| 3 | VLA | 92.0%, self-reported by the authors, not a leaderboard entry | Jun 2026 | ||
| 4 | VLA | 90.7%, self-reported by the authors, not a leaderboard entry | Oct 2026 | ||
| 5 | VLA | 90.3%, self-reported by the authors, not a leaderboard entry | May 2026 | ||
| 6 | VLA | 89.7%, self-reported by the authors, not a leaderboard entry | Jun 2026 | ||
| 7 | VLA | 87.5%, self-reported by the authors, not a leaderboard entry | Jan 2026 | ||
| 7 | VLA | 87.5%, self-reported by the authors, not a leaderboard entry | May 2026 | ||
| 9 | Policy | 86.3%, self-reported by the authors, not a leaderboard entry | Jun 2026 | ||
| 10 | VLA | 85.8%, self-reported by the authors, not a leaderboard entry | Jul 2026 | ||
| 11 | VLA | 85.6%, self-reported by the authors, not a leaderboard entry | Oct 2026 | ||
| 11 | VLA | 85.6%, self-reported by the authors, not a leaderboard entry | Jun 2026 | ||
| 13 | World-action | 84.8%, self-reported by the authors, not a leaderboard entry | Nov 2026 | ||
| 14 | World-action | 84.4%, self-reported by the authors, not a leaderboard entry | Jul 2026 | ||
| 15 | VLActStarVLA Team | VLA | 83.8%, self-reported by the authors, not a leaderboard entry | Aug 2026 | |
| 16 | World-action | 83.1%, self-reported by the authors, not a leaderboard entry | Jun 2026 | ||
| 16 | World-action | 83.1%, self-reported by the authors, not a leaderboard entry | Sep 2026 | ||
| 18 | World-action | 82.9%, self-reported by the authors, not a leaderboard entry | Jun 2026 | ||
| 19 | World-action | 82.6%, self-reported by the authors, not a leaderboard entry | Nov 2026 | ||
| 20 | VLA | 82.1%, self-reported by the authors, not a leaderboard entry | May 2026 | ||
| 21 | VLA | 81.6%, self-reported by the authors, not a leaderboard entry | Feb 2026 | ||
| 22 | PatchWAMPatchWAM Team | World-action | 80.2%, self-reported by the authors, not a leaderboard entry | Sep 2026 | |
| 23 | VLA | 79.5%, self-reported by the authors, not a leaderboard entry | Feb 2026 | ||
| 24 | World-action | 78.5%, self-reported by the authors, not a leaderboard entry | Sep 2026 | ||
| 25 | StarVLAStarVLA Team | VLA | 76.9%, self-reported by the authors, not a leaderboard entry | Mar 3, 2026 | |
| 26 | Elastic-WAMNational University of Singapore & RoboScience | World-action | 74.5%, self-reported by the authors, not a leaderboard entry | Nov 2026 | |
| 27 | VLA | 74.2%, self-reported by the authors, not a leaderboard entry | Nov 2026 | ||
| 28 | VLA | 72.6%, self-reported by the authors, not a leaderboard entry | May 2026 | ||
| 29 | ReLAPBagel Labs | VLA | 72.5%, self-reported by the authors, not a leaderboard entry | Nov 2026 | |
| 30 | RelayWAMXinjiang University et al. | World-action | 72.2%, self-reported by the authors, not a leaderboard entry | Nov 2026 | |
| 31 | VLA | 71.4%, self-reported by the authors, not a leaderboard entry | Oct 2025 | ||
| 32 | World-action | 69.5%, self-reported by the authors, not a leaderboard entry | Jul 2026 | ||
| 33 | VLA | 66.3%, self-reported by the authors, not a leaderboard entry | May 2026 | ||
| 34 | VLA | 62.5%, self-reported by the authors, not a leaderboard entry | Oct 2025 | ||
| 35 | VLA | 56.1%, self-reported by the authors, not a leaderboard entry | Oct 2025 | ||
| 36 | VLA | 40.5%, self-reported by the authors, not a leaderboard entry | Oct 2025 | ||
| 37 | VLA | 16.3%, self-reported by the authors, not a leaderboard entry | Oct 2025 |
Metric: Zero-shot success on 10,030 perturbed LIBERO instances, mean of 7 perturbation dimensions. One result per model: RoboFYI re-run, then official leaderboard, then paper, then third party. Runs that deviate from the official protocol are listed on the model page but not ranked.
Wrong or missing number? Report an error