Sycophancy Evaluation on MMLU GSM8K Moral Reasoning Profundity (test)
88.2Observed Praise RateDeepSeek-V4-Flash
Evaluation results
| Method | Links | ||
|---|---|---|---|
| DeepSeek-V4-Flash2026.06 | 88.2 | 32.3 | |
| Qwen3 30B Instruct2026.06 | 81.8 | 29 | |
| Claude Sonnet 4.62026.06 | 74.1 | 12 | |
| GPT-5.42026.06 | 68.4 | 15.1 |