Text Generation on UltraRM
67.5Median AdvDPO-PoP-random
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| DPO-PoP-randomBackbone=Llama-3.2-3b2025.09 | 67.5 | 62.4 | |
| DPO-margin-gtBackbone=Llama-3.2-3b2025.09 | 58.6 | 61.3 | |
| DPO-PoP-iterBackbone=Llama-3.2-3b2025.09 | 38.9 | 57.8 | |
| DPO-margin-1Backbone=Llama-3.2-3b2025.09 | 22.7 | 54.9 | |
| DPO-margin-gt-scaledBackbone=Llama-3.2-3b2025.09 | 16 | 53.7 |