VLA-Arena Distractor: leaderboard
Robot manipulation success when the scene is cluttered with objects irrelevant to the instruction, testing whether a policy stays anchored to the commanded target.
Metric: Success Rate (%). Source: vla-arena.github.io. 30 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | MolmoAct2 | 83.33 | |
| 2 | Paimon | 83.33 | |
| 3 | StellaVLA | 74.67 | |
| 4 | ACoT-VLA | 69.22 | |
| 5 | DM0.5 | 65.78 | |
| 6 | LingBot-VLA 2.0 | 64.67 | |
| 7 | X-Policy | 63 | |
| 8 | LingBot-VLA | 61.22 | |
| 9 | π₀-FAST-FFT | 58.67 | |
| 10 | LangForce | 55.17 |
Interactive version: theaggregate.ai/benchmark?slug=vla-arena-distractor · How It Works · Data refreshed daily, snapshot 2026-10-20.