Pressure Reveals Character - Corrigibility: Autonomy: leaderboard
Mean 1-5 judge rubric score over 20 multi-turn scenarios testing balancing initiative with deference. One behaviour of the Corrigibility category; a subset of the headline's transcripts.
Metric: Mean judge rubric score over 20 scenarios (1 = failure, 5 = exemplary). Source: storage.googleapis.com. 24 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 4.5 | 4.5 | #156 |
| 2 | Claude Sonnet 4.5 | 4.4 | #250 |
| 3 | GPT-5.2 | 4.4 | #176 |
| 4 | Kimi K2 | 4.25 | #439 |
| 5 | Qwen 3 235B A22B | 4.2 | #623 |
| 6 | Claude Sonnet 4 | 4.15 | #400 |
| 7 | Gemini 2.5 Pro | 4.1 | #266 |
| 8 | Gemini 3 Pro | 4.05 | #137 |
| 9 | Gemini 3 Flash | 4 | #164 |
| 10 | GPT-5 Mini | 4 | #348 |
| 11 | Qwen 3 32B | 4 | #960 |
| 12 | Qwen 3 14B | 4 | #1233 |
| 13 | Mistral Small 3.2 | 3.95 | #1591 |
| 14 | DeepSeek V3.2 | 3.9 | #406 |
| 15 | Mistral Large 3 | 3.85 | #817 |
Interactive version: theaggregate.ai/benchmark?slug=pressure-reveals-character-corrigibility-autonomy · How It Works · Data refreshed daily, snapshot 2026-11-13.