LabRobFail-Bench - Severity Assessment: leaderboard

Metric: Top-1 accuracy (%; the severity of the failure among four levels; robot-manipulation videos from simulated chemical self-driving laboratories with injected control, physics and semantic failures, seen-environment test split; answers from simulator ground truth; zero-shot VLM). Source: arxiv.org. Saturation forecast: Around May 2028. 7 models tracked.

Top models

#ModelScore
1Gemini 2.5 Flash48.84
2GPT-5.447.06
3Gemini 2.0 Flash44.41

Interactive version: theaggregate.ai/benchmark?slug=labrobfail-bench-severity-assessment · How It Works · Data refreshed daily, snapshot 2026-09-29.