Graph-to-Sequence Generation on WikiOFGraph (test)
65.8BLEUT5 (L)
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| T5 (L)evaluation paradigm=Finetune, #P=770M2026.04 | 65.8 | 80.7 | 51.6 | |
| DLM4G-2.oevaluation paradigm=Pretrain, #P=63M2026.04 | 65.4 | 84.4 | 79.1 | |
| DLM4G-2.oevaluation paradigm=Finetune, #P=63M2026.04 | 65.4 | 84.4 | 79.1 | |
| DLM4G-2.oevaluation paradigm=Zero-shot, #P=63M2026.04 | 65.4 | 84.4 | 79.1 | |
| GPT-o4-minievaluation paradigm=Zero-shot, #P=8B2026.04 | 64.8 | 84.7 | 78.3 | |
| DeepSeekevaluation paradigm=Zero-shot, #P=7B2026.04 | 63.3 | 80.9 | 75.2 | |
| LLaMa-3evaluation paradigm=Zero-shot, #P=8B2026.04 | 62.2 | 80.1 | 78.1 | |
| Qwen2.5evaluation paradigm=Zero-shot, #P=7B2026.04 | 62.2 | 68.1 | 74.3 | |
| DLM4G-1.oevaluation paradigm=Pretrain, #P=50M2026.04 | 61.9 | 82.3 | 68.8 | |
| T5 (S)evaluation paradigm=Finetune, #P=60M2026.04 | 38.5 | 68.8 | 47.1 | |
| GPT-2 (B)evaluation paradigm=Finetune, #P=355M2026.04 | 28.5 | 57.2 | 49 | |
| GPT-2 (S)evaluation paradigm=Finetune, #P=124M2026.04 | 16.6 | 42.8 | 48.7 |