SafeSceneReason - Comparative Reasoning: leaderboard
Accuracy on the comparative questions of SafeSceneReason (2026), an industrial-safety multiple-choice benchmark. Percent, higher is better.
Metric: Accuracy (%; the 108 comparative reasoning questions; single-answer multiple choice on the 1,000-question test set balanced across answer positions, built from workplace scene graphs and NIOSH FACE and CSB accident-report figures; temperature 0, unparseable answers counted wrong). Source: arxiv.org. Saturation forecast: Around February 2027. 8 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.5 | 86.11 | #48 |
| 2 | GPT-5.6 Sol | 80.56 | #29 |
| 3 | Gemini 3.5 Flash | 78.7 | #98 |
| 4 | Gemini 3.1 Pro (Preview) | 76.85 | #91 |
| 5 | Kimi K2.6 | 71.3 | #173 |
| 6 | Claude Fable 5 | 67.59 | #26 |
| 7 | Qwen 3.6 27B | 26.85 | #339 |
| 8 | Qwen 3.5 9B | 25 | #803 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=safescenereason-comparative-reasoning · How It Works · Data refreshed daily, snapshot 2026-11-12.