The Compliance Trap: leaderboard
As frontier AI models are deployed in high-stakes decision pipelines, their ability to maintain metacognitive stability (knowing what they do not know, detecting errors, seeking clarification) under adversarial pressure is a critical safety requirement. Current safety evaluations focus on detecting strategic deception (scheming); we investigate a more fundamental failure mode: cognitive collapse.
Metric: Δ (degradation) (self-reported). Source: benchmarklist.com. 9 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Sonnet 4.6 | 1.3 | #90 |
| 2 | GPT-5.4 | 11.9 | #79 |
| 3 | Gemma 3 27B | 14.9 | #607 |
| 4 | Gemini 3.1 Pro (Preview) | 16.2 | #58 |
| 5 | Llama 3.3 70B | 17.8 | #579 |
| 6 | DeepSeek V3.2 | 19.2 | #205 |
| 7 | Grok 4.20 | 20.8 | #175 |
| 8 | DeepSeek V4 Pro | 30.2 | #102 |
Interactive version: theaggregate.ai/benchmark?slug=the-compliance-trap · How It Works · Data refreshed daily, snapshot 2026-10-20.