Language Generation on E2E
70.5BLEULMNet
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| LMNetBackbone=GPT2-M2025.05 | 70.5 | 8.85 | 46.5 | 71.5 | 2.48 | 1.6 | |
| PreLayer*Backbone=GPT2-M2025.05 | 69.7 | 8.81 | 46.1 | 71.4 | 2.49 | 2.7 | |
| AdapterL (11.09M)*Backbone=GPT2-M, Number of Parameters=11.09M2025.05 | 68.9 | 8.71 | 46.1 | 71.3 | 2.47 | 3.9 | |
| LoRABackbone=GPT2-M2025.05 | 68.9 | 8.68 | 46.5 | 71.5 | 2.51 | 2.3 | |
| FT*Backbone=GPT2-M2025.05 | 68.2 | 8.62 | 46.2 | 71 | 2.47 | 4.5 | |
| FTTop2*Backbone=GPT2-M2025.05 | 68.1 | 8.59 | 46 | 70.8 | 2.41 | 6.3 | |
| AdapterH*Backbone=GPT2-M2025.05 | 67.3 | 8.5 | 46 | 70.7 | 2.44 | 6.7 | |
| AdapterL (0.37M)*Backbone=GPT2-M, Number of Parameters=0.37M2025.05 | 66.3 | 8.41 | 45 | 69.8 | 2.4 | 8 | |
| No AdaptationBackbone=GPT2-M2025.05 | 0 | 0.42 | 0.04 | 0.16 | 0 | 9 |