LLM Alignment on TL;DR (test)
68.56Win Rate (GPT-4o)DPPrefSyn
Evaluation Results
| Method | Links | |
|---|---|---|
| DPPrefSynData Type=Synthetic, Preference Alignment=SFT+DPO, Privacy Budget (epsilon)=4, Downstream Model=Pythia-2.8B2026.05 | 68.56 | |
| DP-FTData Type=Original, Preference Alignment=SFT+DPO, Privacy Budget (epsilon)=4, Downstream Model=Pythia-2.8B2026.05 | 63 | |
| DPPrefSynData Type=Synthetic, Preference Alignment=SFT, Privacy Budget (epsilon)=4, Downstream Model=Pythia-2.8B2026.05 | 62.72 | |
| DP-FTData Type=Original, Preference Alignment=SFT, Privacy Budget (epsilon)=4, Downstream Model=Pythia-2.8B2026.05 | 22.45 |