Next Token Prediction on WildChat
51Next Token AccuracyDP-FT
Evaluation results
| Method | Links | |
|---|---|---|
| DP-FTData Type=Private, Size=2,000, Backbone=GPT-2, Privacy Budget (epsilon)=inf2026.02 | 51 | |
| DP-FTData Type=Private, Size=2,000, Backbone=GPT-2, Privacy Budget (epsilon)=42026.02 | 47.95 | |
| DP-FTData Type=Private, Size=2,000, Backbone=GPT-2, Privacy Budget (epsilon)=22026.02 | 47.59 | |
| DP-FTData Type=Private, Size=2,000, Backbone=GPT-2, Privacy Budget (epsilon)=12026.02 | 47.06 | |
| DP-RFTData Type=Synthetic, Size=2,000, Backbone=GPT-2, Privacy Budget (epsilon)=42026.02 | 45.42 | |
| DP-RFTData Type=Synthetic, Size=2,000, Backbone=GPT-2, Privacy Budget (epsilon)=22026.02 | 45.3 | |
| DP-RFTData Type=Synthetic, Size=2,000, Backbone=GPT-2, Privacy Budget (epsilon)=inf2026.02 | 45.27 | |
| DP-RFTData Type=Synthetic, Size=2,000, Backbone=GPT-2, Privacy Budget (epsilon)=12026.02 | 45.03 | |
| Aug-PEData Type=Synthetic, Size=2,000, Backbone=GPT-2, Privacy Budget (epsilon)=22026.02 | 43.65 | |
| Aug-PEData Type=Synthetic, Size=2,000, Backbone=GPT-2, Privacy Budget (epsilon)=42026.02 | 43.61 | |
| Aug-PEData Type=Synthetic, Size=2,000, Backbone=GPT-2, Privacy Budget (epsilon)=12026.02 | 43.46 | |
| Aug-PEData Type=Synthetic, Size=2,000, Backbone=GPT-2, Privacy Budget (epsilon)=inf2026.02 | 42.39 | |
| QWENData Type=Synthetic, Size=2,000, Backbone=GPT-2, Privacy Budget (epsilon)=inf2026.02 | 33 | |
| QWENData Type=Synthetic, Size=2,000, Backbone=GPT-2, Privacy Budget (epsilon)=42026.02 | 33 | |
| QWENData Type=Synthetic, Size=2,000, Backbone=GPT-2, Privacy Budget (epsilon)=22026.02 | 33 | |
| QWENData Type=Synthetic, Size=2,000, Backbone=GPT-2, Privacy Budget (epsilon)=12026.02 | 33 | |
| DP-FTData Type=Private, Size=2,000, Backbone=BERTSmall, Privacy Budget (epsilon)=inf2026.02 | 19.62 | |
| DP-RFTData Type=Synthetic, Size=2,000, Backbone=BERTSmall, Privacy Budget (epsilon)=22026.02 | 14.14 | |
| DP-RFTData Type=Synthetic, Size=2,000, Backbone=BERTSmall, Privacy Budget (epsilon)=42026.02 | 14.08 | |
| Aug-PEData Type=Synthetic, Size=2,000, Backbone=BERTSmall, Privacy Budget (epsilon)=42026.02 | 14.04 | |
| DP-RFTData Type=Synthetic, Size=2,000, Backbone=BERTSmall, Privacy Budget (epsilon)=inf2026.02 | 13.93 | |
| DP-RFTData Type=Synthetic, Size=2,000, Backbone=BERTSmall, Privacy Budget (epsilon)=12026.02 | 13.89 | |
| Aug-PEData Type=Synthetic, Size=2,000, Backbone=BERTSmall, Privacy Budget (epsilon)=22026.02 | 13.81 | |
| Aug-PEData Type=Synthetic, Size=2,000, Backbone=BERTSmall, Privacy Budget (epsilon)=12026.02 | 13.63 | |
| Aug-PEData Type=Synthetic, Size=2,000, Backbone=BERTSmall, Privacy Budget (epsilon)=inf2026.02 | 13.29 | |
| DP-FTData Type=Private, Size=2,000, Backbone=BERTSmall, Privacy Budget (epsilon)=42026.02 | 8.28 | |
| DP-FTData Type=Private, Size=2,000, Backbone=BERTSmall, Privacy Budget (epsilon)=22026.02 | 7.37 | |
| DP-FTData Type=Private, Size=2,000, Backbone=BERTSmall, Privacy Budget (epsilon)=12026.02 | 5.3 | |
| QWENData Type=Synthetic, Size=2,000, Backbone=BERTSmall, Privacy Budget (epsilon)=inf2026.02 | 5.22 | |
| QWENData Type=Synthetic, Size=2,000, Backbone=BERTSmall, Privacy Budget (epsilon)=42026.02 | 5.22 | |
| QWENData Type=Synthetic, Size=2,000, Backbone=BERTSmall, Privacy Budget (epsilon)=22026.02 | 5.22 | |
| QWENData Type=Synthetic, Size=2,000, Backbone=BERTSmall, Privacy Budget (epsilon)=12026.02 | 5.22 |