HELM Robo-Reward-Bench - Cmu Play Fusion: leaderboard
The "Cmu Play Fusion" subtask of HELM Robo-Reward-Bench. Stanford HELM's Robo-Reward-Bench: vision-language models act as reward models over robot manipulation episodes, predicting task progress or success from video.
Metric: Absolute Error. Source: crfm.stanford.edu. Saturation forecast: Around September 2027. 25 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Pro (Preview) | 0.63 | #67 |
| 2 | Gemini 3 Flash (Preview) | 0.71 | #86 |
| 3 | Gemini 2.5 Flash | 0.78 | #247 |
| 4 | Gemini 2.5 Pro | 0.8 | #155 |
| 5 | GPT-5.5 (2026-04-23) | 0.84 | #23 |
| 6 | GPT-5.5 (xHigh) | 0.97 | #30 (GPT-5.5) |
| 7 | GPT-5 | 0.99 | #101 |
| 8 | GPT-5.1 (2025-11-13) | 1 | #106 |
| 9 | Gemini 2.5 Flash Lite | 1.07 | #426 |
| 10 | GPT-5 Mini (2025-08-07) | 1.14 | #175 |
| 11 | GPT-5.2 | 1.18 | #114 |
| 12 | Llama 4 Maverick Instruct | 1.28 | #460 |
| 13 | Llama 4 Scout Instruct | 1.37 | #541 |
| 14 | GPT-5 Nano | 1.51 | #430 |
No result here: #4 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=helm-robo-reward-bench-cmu-play-fusion · How It Works · Data refreshed daily, snapshot 2026-10-24.