Domain Mixture Recovery on Average (SNLI, AGNews, Yelp, Yahoo)
0.005Average MSETIES
Evaluation Results
| Method | Links | |
|---|---|---|
| TIESModel Architecture=BERT, Method Category=Supervised 2-Layer MLP2026.07 | 0.005 | |
| LERPModel Architecture=BERT, Method Category=Supervised 2-Layer MLP2026.07 | 0.006 | |
| Sample-level MIModel Architecture=BERT, Method Category=Baselines2026.07 | 0.007 | |
| LERPModel Architecture=BERT, Method Category=Unsupervised Softmax2026.07 | 0.013 | |
| TIESModel Architecture=BERT, Method Category=Unsupervised Softmax2026.07 | 0.013 | |
| Real TrajectoryModel Architecture=BERT, Method Category=Supervised 2-Layer MLP2026.07 | 0.014 | |
| LERPModel Architecture=GPT-2-Small, Method Category=Supervised 2-Layer MLP2026.07 | 0.023 | |
| TIESModel Architecture=GPT-2-Small, Method Category=Supervised 2-Layer MLP2026.07 | 0.025 | |
| Real TrajectoryModel Architecture=GPT-2-Small, Method Category=Supervised 2-Layer MLP2026.07 | 0.03 | |
| SLERPModel Architecture=BERT, Method Category=Unsupervised Softmax2026.07 | 0.032 | |
| TIESModel Architecture=GPT-2-Small, Method Category=Unsupervised Softmax2026.07 | 0.032 | |
| SLERPModel Architecture=BERT, Method Category=Supervised 2-Layer MLP2026.07 | 0.034 | |
| LERPModel Architecture=GPT-2-Small, Method Category=Unsupervised Softmax2026.07 | 0.035 | |
| Centroid GuessModel Architecture=BERT, Method Category=Baselines2026.07 | 0.04 | |
| Centroid GuessModel Architecture=GPT-2-Small, Method Category=Baselines2026.07 | 0.04 | |
| Sample-level MIModel Architecture=GPT-2-Small, Method Category=Baselines2026.07 | 0.042 | |
| Real TrajectoryModel Architecture=GPT-2-Small, Method Category=Unsupervised Softmax2026.07 | 0.048 | |
| SLERPModel Architecture=GPT-2-Small, Method Category=Supervised 2-Layer MLP2026.07 | 0.055 | |
| Random GuessModel Architecture=BERT, Method Category=Baselines2026.07 | 0.079 | |
| Random GuessModel Architecture=GPT-2-Small, Method Category=Baselines2026.07 | 0.079 | |
| Real TrajectoryModel Architecture=BERT, Method Category=Unsupervised Softmax2026.07 | 0.097 | |
| SLERPModel Architecture=GPT-2-Small, Method Category=Unsupervised Softmax2026.07 | 0.13 |