Safety assessment on StrongReject
0.179Personalization Bias (PB)Identity-Robust Generation
Evaluation Results
| Method | Links | |
|---|---|---|
| Identity-Robust GenerationModel=Llama3.3-70B2026.01 | 0.179 | |
| Identity-Robust GenerationModel=Qwen32026.01 | 0.201 | |
| Prompt SteeringModel=Qwen32026.01 | 0.282 | |
| Identity-Robust GenerationModel=gpt-oss-20B2026.01 | 0.325 | |
| Vanilla GenerationModel=Llama3.3-70B2026.01 | 0.384 | |
| Vanilla GenerationModel=Qwen32026.01 | 0.408 | |
| Prompt SteeringModel=Llama3.3-70B2026.01 | 0.444 | |
| Vanilla GenerationModel=gpt-oss-20B2026.01 | 0.831 | |
| Prompt SteeringModel=gpt-oss-20B2026.01 | 1.413 |