Pressure Reveals Character - Corrigibility: Appropriate Autonomy: leaderboard
Mean 1-5 judge rubric score over 6 multi-turn scenarios testing knowing when to act and when to ask for guidance. One behaviour of the Corrigibility category; a subset of the headline's transcripts.
Metric: Mean judge rubric score over 6 scenarios (1 = failure, 5 = exemplary). Source: storage.googleapis.com. 24 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Sonnet 4.5 | 4.83 | #250 |
| 2 | Claude Opus 4.5 | 4.83 | #156 |
| 3 | GPT-5.2 | 4.5 | #176 |
| 4 | Gemini 3 Pro | 4.17 | #137 |
| 5 | Claude Sonnet 4 | 4.17 | #400 |
| 6 | Gemini 2.5 Pro | 3.83 | #266 |
| 7 | Gemini 2.5 Flash | 3.83 | #453 |
| 8 | GPT-OSS-120B | 3.83 | #700 |
| 9 | DeepSeek V3.2 | 3.83 | #406 |
| 10 | Llama 4 Maverick | 3.83 | #1099 |
| 11 | Mistral Medium 3.1 | 3.83 | #852 |
| 12 | Gemini 3 Flash | 3.67 | #164 |
| 13 | Kimi K2 | 3.67 | #439 |
| 14 | Llama 3.1 405B Instruct | 3.67 | #1115 |
| 15 | Qwen 3 235B A22B | 3.67 | #623 |
Interactive version: theaggregate.ai/benchmark?slug=pressure-reveals-character-corrigibility-appropriate-autonomy · How It Works · Data refreshed daily, snapshot 2026-11-13.