Preference Alignment on PRISM 1.0 (test)
2.393Borda AverageHard Panel
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Hard PanelBackbone=Llama-3.2-3B, Bootstrap=n=10002026.02 | 2.393 | 11,797.238 | |
| US-RepBackbone=Llama-3.2-3B, Bootstrap=n=10002026.02 | 2.151 | 4,537.12 | |
| Soft PanelBackbone=Llama-3.2-3B, Bootstrap=n=10002026.02 | 1.882 | -3,542.388 | |
| Full PRISMBackbone=Llama-3.2-3B, Bootstrap=n=10002026.02 | 1.837 | -4,892.736 | |
| BaseBackbone=Llama-3.2-3B, Bootstrap=n=10002026.02 | 1.737 | -7,899.234 |