HELM Robo-Reward-Bench - Nyu Door Opening Surprising Effectiveness: leaderboard
The "Nyu Door Opening Surprising Effectiveness" 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 2.5 Pro | 0.74 | #155 |
| 2 | GPT-5 | 1.07 | #101 |
| 3 | GPT-4.1 (2025-04-14) | 1.09 | #252 |
| 4 | Gemini 2.0 Flash Lite | 1.12 | #455 |
| 5 | GPT-4.1 Mini | 1.2 | #358 |
| 6 | O1 (2024-12-17) | 1.2 | #153 |
| 7 | Gemini 2.5 Flash | 1.25 | #247 |
| 8 | GPT-5 Mini (2025-08-07) | 1.26 | #175 |
| 9 | Gemini 2.0 Flash | 1.34 | #345 |
| 10 | GPT-4o (2024-11-20) | 1.35 | #379 |
| 11 | Gemini 2.5 Flash Lite | 1.66 | #426 |
| 12 | GPT-5 Nano | 1.84 | #430 |
| 13 | Qwen2.5-Omni-7B | 3.63 | #634 |
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-nyu-door-opening-surprising-effectiveness · How It Works · Data refreshed daily, snapshot 2026-10-24.