Sycophancy Evaluation on TruthfulQA Adversarial n=50
2Sycophantic Responses Count
Gemini 2.5 Flash (Static Guardrails)
Updated 17d ago
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| 2 | 4 | 0.5 | 91.3 | ||
2026.04 | 7 | 14 | 5.8 | 69.6 | |
2026.04 | 23 | 46 | 31.8 | - |
Gemini 2.5 Flash (Static Guardrails)
| Method | Links | ||||
|---|---|---|---|---|---|
| 2 | 4 | 0.5 | 91.3 | ||
2026.04 | 7 | 14 | 5.8 | 69.6 | |
2026.04 | 23 | 46 | 31.8 | - |