Factuality assessment on TruthfulQA using Personalization Bias (PB)
0.148Personalization Bias (PB)Identity-Robust Generation
Evaluation Results
| Method | Links | |
|---|---|---|
| Identity-Robust GenerationModel=Llama3.3-70B2026.01 | 0.148 | |
| Identity-Robust GenerationModel=Qwen32026.01 | 0.241 | |
| Identity-Robust GenerationModel=gpt-oss-20B2026.01 | 0.346 | |
| Prompt SteeringModel=Qwen32026.01 | 0.625 | |
| Prompt SteeringModel=Llama3.3-70B2026.01 | 0.729 | |
| Vanilla GenerationModel=Qwen32026.01 | 0.864 | |
| Vanilla GenerationModel=Llama3.3-70B2026.01 | 0.894 | |
| Vanilla GenerationModel=gpt-oss-20B2026.01 | 1.504 | |
| Prompt SteeringModel=gpt-oss-20B2026.01 | 1.547 |