HELM Robo-Reward-Bench - Utaustin Mutex: leaderboard
The "Utaustin Mutex" 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 December 2026. 20 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 2.0 Flash | 1.02 | #344 |
| 2 | Gemini 2.5 Pro | 1.06 | #153 |
| 3 | O1 (2024-12-17) | 1.13 | #152 |
| 4 | GPT-5 Mini (2025-08-07) | 1.16 | #174 |
| 5 | Gemini 2.5 Flash | 1.3 | #242 |
| 6 | GPT-4.1 Mini | 1.33 | #351 |
| 7 | GPT-4o (2024-11-20) | 1.41 | #378 |
| 8 | Gemini 2.0 Flash Lite | 1.46 | #453 |
| 9 | GPT-5 | 1.5 | #104 |
| 10 | GPT-5 Nano | 1.55 | #428 |
| 11 | GPT-4.1 (2025-04-14) | 1.69 | #250 |
| 12 | Gemini 2.5 Flash Lite | 1.9 | #421 |
| 13 | Qwen2.5-Omni-7B | 2.86 | #632 |
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-utaustin-mutex · How It Works · Data refreshed daily, snapshot 2026-10-23.