Safety Discovery in Text-based Scenarios on DB Migration Side Effects
49R ScoreReward-Only
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Reward-OnlyModel=Claude Sonnet 4.62026.04 | 49 | 38 | 5 | |
| CoTModel=Claude Sonnet 4.62026.04 | 48 | 37 | 5 | |
| StaticModel=Claude Sonnet 4.62026.04 | 48 | 37 | 5 | |
| Reward-OnlyModel=Gemini 3 Flash2026.04 | 48 | 37 | 5 | |
| CoTModel=Gemini 3 Flash2026.04 | 48 | 37 | 5 | |
| StaticModel=Gemini 3 Flash2026.04 | 48 | 37 | 5 | |
| EPO-SafeModel=Claude Sonnet 4.62026.04 | 46 | 46 | 0 | |
| EPO-SafeModel=Gemini 3 Flash2026.04 | 46 | 46 | 0 |