Struct to Text on CommonGen
64ROUGE-1Supervised Model (T5-11B)
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Supervised Model (T5-11B)mode=supervised2021.09 | 64 | 29.4 | 54.5 | |
| LaMDA-PTprotocol=few-shot, k=32021.09 | 56.7 | 29.6 | 48.5 | |
| FLAN 137Bprotocol=few-shot, template_selection=average, k=16, number_of_templates=62021.09 | 56.6 | 30.9 | 50.7 | |
| FLAN 137Bprotocol=few-shot, template_selection=best dev, k=16, number_of_templates=62021.09 | 56.4 | 29.9 | 51 | |
| FLAN 137Bprotocol=zero-shot, template_selection=best dev, number_of_templates=62021.09 | 56.3 | 27.6 | 48.7 | |
| LOGO (Entropy)Backbone=LLaMA-3.1-8B2025.11 | 56 | 25 | 45.8 | |
| BaseBackbone=LLaMA-3.1-8B2025.11 | 54.8 | 23.6 | 44 | |
| LOGO (Norm)Backbone=Qwen-2.5-7B2025.11 | 54.7 | 23.1 | 44.8 | |
| FLAN 137Bprotocol=zero-shot, template_selection=average, number_of_templates=62021.09 | 54.6 | 28.8 | 48.4 | |
| Adapter SoupBackbone=DeepSeek-LLM-7B-Base2025.11 | 52 | 21.8 | 43.3 | |
| LaMDA-PTprotocol=zero-shot2021.09 | 3.9 | 1.5 | 3.2 |