Correction Suppression Benchmark: leaderboard
LLMs reliably correct false claims when presented in isolation, yet when the same claims are embedded in task-oriented requests, they often comply rather than correct. We term this failure mode correction suppression and construct a benchmark of 300 false premises to systematically evaluate it across eight models. Suppression rates range from 19% to 90%, with four models exceeding 80%, establishin
Metric: Suppression Rate (self-reported). Source: benchmarklist.com. Saturation forecast: Not forecast. 8 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.1 | 89.6 | #135 |
| 2 | DeepSeek V3.2 | 84 | #205 |
| 3 | Gemini 3 Flash | 82.9 | #99 |
| 4 | Grok 4.1 Fast | 81.2 | #217 |
| 5 | Llama 3.1 8B | 47.4 | #1151 |
| 6 | Qwen 3.5 9B | 46 | #379 |
| 7 | Claude Sonnet 4.5 | 31 | #142 |
| 8 | Qwen 3.5 Plus | 19.3 | #125 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=correction-suppression-benchmark · How It Works · Data refreshed daily, snapshot 2026-10-20.