Natural Language Generation on Task
26.7BLEUMultitask Finetuning
Evaluation results
| Method | Links | |
|---|---|---|
| Multitask FinetuningModel=GEMMA-2B, Warm-start weights=Simple Merging2024.12 | 26.7 | |
| Multitask FinetuningModel=GEMMA-2B, Warm-start weights=Hessian Weighted2024.12 | 26.7 | |
| Multitask FinetuningModel=GEMMA-2B, Selection=Global maximum2024.12 | 26.7 | |
| Multitask FinetuningModel=GEMMA-2B, Warm-start weights=Mixture Weighted2024.12 | 26.6 | |
| Mixture WeightedModel=GEMMA-2B2024.12 | 26.2 | |
| Simple MergingModel=GEMMA-2B2024.12 | 25.9 | |
| Hessian WeightedModel=GEMMA-2B2024.12 | 25.9 |