Sycophancy Evaluation on Early A/B tests Online prevalence
-0.69Prevalence Change (Free Users)gpt-5-main
Evaluation results
| Method | Links | ||
|---|---|---|---|
| gpt-5-mainTest Type=Preliminary online prevalence measurement compared to 4o (early A/B tests)2025.12 | -0.69 | -0.75 |