Medical Subject Classification (Chemicals and Drugs) on PubMed Synthetic GPT-2-large generated (test)
87.88AccuracyFederated GPT-2-large
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Federated GPT-2-largeCs%=40%, epsilon=∞, refinement=false2026.03 | 87.88 | 87.88 | |
| Federated GPT-2-largeCs%=30%, epsilon=∞, refinement=false2026.03 | 86.12 | 86.05 | |
| Federated GPT-2-largeCs%=40%, epsilon=8, refinement=true2026.03 | 84.26 | 84.48 | |
| Federated GPT-2-largeCs%=30%, epsilon=8, refinement=true2026.03 | 83.12 | 83.18 | |
| Federated GPT-2-largeCs%=20%, epsilon=8, refinement=true2026.03 | 82.68 | 82.8 | |
| Federated GPT-2-largeCs%=10%, epsilon=8, refinement=true2026.03 | 81 | 80.74 | |
| Federated GPT-2-largeCs%=5%, epsilon=8, refinement=true2026.03 | 80.28 | 80.24 | |
| Federated GPT-2-largeCs%=20%, epsilon=∞, refinement=false2026.03 | 79.68 | 79.13 | |
| Federated GPT-2-largeCs%=10%, epsilon=∞, refinement=false2026.03 | 69.04 | 64.6 | |
| Federated GPT-2-largeCs%=5%, epsilon=∞, refinement=false2026.03 | 66.72 | 61.13 | |
| GPT-2-large (Pretrained)Cs%=Pre., epsilon=8, refinement=true2026.03 | 64.64 | 54.72 | |
| GPT-2-large (Pretrained)Cs%=Pre., epsilon=∞, refinement=false2026.03 | 64.56 | 58.81 | |
| Federated GPT-2-largeCs%=40%, epsilon=8, refinement=false2026.03 | 64 | 57.65 | |
| Federated GPT-2-largeCs%=30%, epsilon=8, refinement=false2026.03 | 63.8 | 55.77 | |
| Federated GPT-2-largeCs%=10%, epsilon=8, refinement=false2026.03 | 63.32 | 52.03 | |
| Federated GPT-2-largeCs%=20%, epsilon=8, refinement=false2026.03 | 63.04 | 53.46 | |
| Federated GPT-2-largeCs%=5%, epsilon=8, refinement=false2026.03 | 62.2 | 50.19 |