RLHF on HH-RLHF (test)
88.94Average ScoreGEOALIGN
Evaluation Results
| Method | Links | |
|---|---|---|
| GEOALIGNBackbone=Qwen3-4B2026.06 | 88.94 | |
| GEOALIGNBackbone=Qwen3-1.7B2026.06 | 88.85 | |
| PF-PPO (BR)Backbone=Qwen3-4B2026.06 | 87.71 | |
| PARBackbone=Qwen3-4B2026.06 | 87.61 | |
| SEED-GRPOBackbone=Qwen3-4B2026.06 | 87.43 | |
| BASE-GRPOBackbone=Qwen3-4B2026.06 | 86.72 | |
| PF-PPO (BW)Backbone=Qwen3-4B2026.06 | 86.6 | |
| PODSBackbone=Qwen3-4B2026.06 | 85.94 | |
| PF-PPO (BR)Backbone=Qwen3-1.7B2026.06 | 85.56 | |
| PARBackbone=Qwen3-1.7B2026.06 | 84.91 | |
| SEED-GRPOBackbone=Qwen3-1.7B2026.06 | 84.24 | |
| PF-PPO (BW)Backbone=Qwen3-1.7B2026.06 | 83.96 | |
| BASE-GRPOBackbone=Qwen3-1.7B2026.06 | 83.54 | |
| PODSBackbone=Qwen3-1.7B2026.06 | 83.39 | |
| SFTBackbone=Qwen3-4B2026.06 | 44.73 | |
| SFTBackbone=Qwen3-1.7B2026.06 | 43.05 |