RoboAbstention: leaderboard
RoboAbstention (2026, The Yes-Man Syndrome) gives vision-language robot planners 6,069 image and instruction pairs that all call for abstaining or asking for clarification: missing or ambiguous referents, subjective or underspecified intent, infeasible, contradictory or false-premise requests. The score is the abstention rate in percent; higher is better.
Metric: Abstention rate (%) over all 6,069 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. 8 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Setting | Score | Overall rank |
|---|---|---|---|---|
| 1 | Gemini 2.5 Flash | 39 | #463 | |
| 2 | Claude Sonnet 4.6 | 26.4 | #173 | |
| 3 | Llama 4 Maverick | 18.8 | #1110 | |
| 4 | Gemini Robotics-ER 1.6 | 16.5 | #170 | |
| 5 | GPT-5.4 Mini | Low | 14.8 | #406 |
| 6 | GPT-5.4 | 12.9 | #135 | |
| 7 | GPT-5.4 Nano | 12.5 | #651 | |
| 8 | Qwen 3.5 27B | 9.2 | #375 |
Interactive version: theaggregate.ai/benchmark?slug=roboabstention · How It Works · Data refreshed daily, snapshot 2026-11-21.