Pairwise Ranking on Seven Harm Categories
83.1Insult Pairwise ScoreOAI-RM
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| OAI-RMBase Model=Llama-3.1-8B2025.03 | 83.1 | 81.9 | 79.7 | 80.5 | 82.3 | 78.9 | 80.1 | |
| Shadow-RMBase Model=Llama-3.1-8B2025.03 | 9.3 | 9.9 | 10.5 | 10.4 | 9.4 | 10.7 | 10.7 | |
| Anthropic-RMBase Model=Llama-3.1-8B2025.03 | 2.5 | 2.3 | 2.4 | 2.9 | 2.4 | 3.1 | 3.7 |