Preference Evaluation on Anthropic-SafeRLHF
41.7Win Rateπbias (rubric-based preference attack)
Evaluation Results
| Method | Links | |
|---|---|---|
| πbias (rubric-based preference attack)Comparison=πbias vs. πseed, Training setting=Target-only (T), Evaluator model=LLaMA-3-8B, Decoding strategy=Best-of-42026.02 | 41.7 | |
| πbias (rubric-based preference attack)Comparison=πbias vs. πseed, Training setting=Benchmark+Target (BT), Evaluator model=LLaMA-3-8B, Decoding strategy=Best-of-42026.02 | 34.1 |