Preference Modeling on BT-guided
1.09NLLPPT (Distribution) + TS
Evaluation Results
| Method | Links | |
|---|---|---|
| PPT (Distribution) + TSBackbone=LLAMA3-8B, Temperature Scaling=true2026.05 | 1.09 | |
| PPT (Distribution) + TSBackbone=QWEN2-7B, Temperature Scaling=true2026.05 | 1.09 | |
| PPT (Distribution)Backbone=LLAMA3-8B, Temperature Scaling=false2026.05 | 1.13 | |
| PPT (Distribution)Backbone=QWEN2-7B, Temperature Scaling=false2026.05 | 1.28 | |
| PPT (Mean)Backbone=LLAMA3-8B, Temperature Scaling=false2026.05 | 1.37 | |
| PPT (Mean)Backbone=QWEN2-7B, Temperature Scaling=false2026.05 | 1.77 | |
| BaseBackbone=LLAMA3-8B, Temperature Scaling=false2026.05 | 2.12 | |
| BaseBackbone=QWEN2-7B, Temperature Scaling=false2026.05 | 2.21 |