RoboAbstention - Physical Infeasibility: leaderboard
The Physical Infeasibility category of RoboAbstention (2026, The Yes-Man Syndrome): 50 instructions that the robot cannot physically carry out, given with an image to a vision-language robot planner that should abstain or ask. The score is the abstention rate in percent; higher is better.
Metric: Abstention rate (%) on the 50 Physical Infeasibility instructions, image plus instruction to an embodied VLM planner, every instruction warranting abstention, act or abstain judged by GPT-5.4 mini (97.5 percent agreement with human labels), API default sampling; higher is better. Source: arxiv.org. Saturation forecast: Not forecast. 8 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Setting | Score | Overall rank |
|---|---|---|---|---|
| 1 | Gemini 2.5 Flash | 26 | #463 | |
| 2 | Qwen 3.5 27B | 10 | #375 | |
| 3 | Llama 4 Maverick | 8 | #1110 | |
| 4 | GPT-5.4 Mini | High | 8 | #406 |
| 5 | Gemini Robotics-ER 1.6 | 6 | #170 | |
| 6 | Claude Sonnet 4.6 | 4 | #173 | |
| 7 | GPT-5.4 | 2 | #135 | |
| 8 | GPT-5.4 Nano | 2 | #651 |
No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=roboabstention-physical-infeasibility · How It Works · Data refreshed daily, snapshot 2026-11-21.