Discriminative Performance on HH_Golden 42.5k (test)
99.61AccuracyFA-DPO
Evaluation Results
| Method | Links | |
|---|---|---|
| FA-DPOFlipping ratio (η)=0%, Model Backbone=Pythia-1B2025.11 | 99.61 | |
| cDPOFlipping ratio (η)=0%, Model Backbone=Pythia-1B2025.11 | 99.24 | |
| FA-DPOFlipping ratio (η)=10%, Model Backbone=Pythia-1B2025.11 | 99.17 | |
| FA-DPOFlipping ratio (η)=30%, Model Backbone=Pythia-1B2025.11 | 99.02 | |
| DPOFlipping ratio (η)=0%, Model Backbone=Pythia-1B2025.11 | 98.89 | |
| FA-DPOFlipping ratio (η)=40%, Model Backbone=Pythia-1B2025.11 | 98.83 | |
| ROPOFlipping ratio (η)=0%, Model Backbone=Pythia-1B2025.11 | 98.8 | |
| SIMPOFlipping ratio (η)=0%, Model Backbone=Pythia-1B2025.11 | 98.27 | |
| FA-DPOFlipping ratio (η)=20%, Model Backbone=Pythia-1B2025.11 | 98.1 | |
| rDPOFlipping ratio (η)=0%, Model Backbone=Pythia-1B2025.11 | 96.8 | |
| rDPOFlipping ratio (η)=10%, Model Backbone=Pythia-1B2025.11 | 96.67 | |
| SIMPOFlipping ratio (η)=10%, Model Backbone=Pythia-1B2025.11 | 93.63 | |
| DPOFlipping ratio (η)=10%, Model Backbone=Pythia-1B2025.11 | 93.5 | |
| ROPOFlipping ratio (η)=10%, Model Backbone=Pythia-1B2025.11 | 93.27 | |
| cDPOFlipping ratio (η)=10%, Model Backbone=Pythia-1B2025.11 | 93.2 | |
| rDPOFlipping ratio (η)=20%, Model Backbone=Pythia-1B2025.11 | 87.83 | |
| DPOFlipping ratio (η)=20%, Model Backbone=Pythia-1B2025.11 | 83.53 | |
| ROPOFlipping ratio (η)=20%, Model Backbone=Pythia-1B2025.11 | 82.87 | |
| SIMPOFlipping ratio (η)=20%, Model Backbone=Pythia-1B2025.11 | 82.83 | |
| cDPOFlipping ratio (η)=20%, Model Backbone=Pythia-1B2025.11 | 82.5 | |
| rDPOFlipping ratio (η)=30%, Model Backbone=Pythia-1B2025.11 | 73.73 | |
| SIMPOFlipping ratio (η)=30%, Model Backbone=Pythia-1B2025.11 | 73.33 | |
| DPOFlipping ratio (η)=30%, Model Backbone=Pythia-1B2025.11 | 73.3 | |
| cDPOFlipping ratio (η)=30%, Model Backbone=Pythia-1B2025.11 | 73.3 | |
| ROPOFlipping ratio (η)=30%, Model Backbone=Pythia-1B2025.11 | 72.33 | |
| ROPOFlipping ratio (η)=40%, Model Backbone=Pythia-1B2025.11 | 62 | |
| SIMPOFlipping ratio (η)=40%, Model Backbone=Pythia-1B2025.11 | 60.93 | |
| cDPOFlipping ratio (η)=40%, Model Backbone=Pythia-1B2025.11 | 60.87 | |
| DPOFlipping ratio (η)=40%, Model Backbone=Pythia-1B2025.11 | 58.63 | |
| rDPOFlipping ratio (η)=40%, Model Backbone=Pythia-1B2025.11 | 56.8 |