RoboLab-120 leaderboard
NVIDIA SRL benchmark of 120 held-out tasks (65 simple, 38 moderate, 18 complex) in which policies trained on real-world data run in simulation without sim co-training. Leaderboard submissions are currently paused.
Simulation · Franka (DROID-style, inferred) · Isaac Sim 5.0 / Isaac Lab 2.2Report an error
86.0%Best · AMP1
24Models ranked
OpenSaturation
Nov 18, 2026Numbers checked
| # | Model | Type | Score | Date | Source |
|---|---|---|---|---|---|
| 1 | Agent system | 86.0% | Nov 17, 2026 | ||
| 2 | Agent system | 57.1% | Nov 17, 2026 | ||
| 3 | World-action | 51.4%, self-reported by the authors, not a leaderboard entry | Nov 16, 2026 | ||
| 4 | LakesidePrI | World-action | 51.0% | Nov 17, 2026 | |
| 5 | VLA | 49.2% | Nov 17, 2026 | ||
| 6 | World-action | 47.4%, self-reported by the authors, not a leaderboard entry | Nov 2026 | ||
| 7 | World-action | 46.2% | Nov 17, 2026 | ||
| 8 | World-action | 42.9% | Nov 17, 2026 | ||
| 9 | Agent system | 42.3% | Nov 17, 2026 | ||
| 10 | HiDream-O1-EmbodiedHiDream.ai | VLA | 39.9% | Nov 17, 2026 | |
| 11 | World-action | 39.6% | Nov 17, 2026 | ||
| 12 | World-action | 39.0% | Nov 17, 2026 | ||
| 13 | World-action | 36.8% | Nov 17, 2026 | ||
| 14 | Agent system | 34.4% | Nov 17, 2026 | ||
| 15 | VLA | 33.3% | Nov 17, 2026 | ||
| 16 | Agent system | 28.2% | Nov 17, 2026 | ||
| 17 | VLA | 28.0% | Nov 17, 2026 | ||
| 18 | World-action | 25.7% | Nov 17, 2026 | ||
| 19 | World-action | 22.9% | Nov 17, 2026 | ||
| 20 | VLA | 15.5% | Nov 17, 2026 | ||
| 21 | VLA | 8.6%, self-reported by the authors, not a leaderboard entry | Nov 17, 2026 | ||
| 22 | VLA | 7.2% | Nov 17, 2026 | ||
| 23 | VLA | 5.0% | Nov 17, 2026 | ||
| 24 | VLA | 3.4% | Nov 17, 2026 |
Metric: Success rate over 120 held-out tasks (10 episodes per task; some agentic entries ran 25). 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