RDF-to-text generation on GEM2 Fictional (test)
93.7GrammaticalityLlama 3.3
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Llama 3.3Architecture=Neural model2025.12 | 93.7 | 1.8 | 9.6 | |
| GPT-4.1Architecture=Rule-based NLG2025.12 | 73.8 | 3.6 | 9.8 | |
| Qwen 3 235BArchitecture=Rule-based NLG2025.12 | 63.2 | 4.3 | 6.6 | |
| BARTArchitecture=Neural model2025.12 | 61.9 | 58 | 59.9 | |
| Qwen 2.5 72BArchitecture=Rule-based NLG2025.12 | 60.3 | 3 | 9.8 | |
| Llama 3.3 70BArchitecture=Rule-based NLG2025.12 | 55.1 | 6.5 | 20.8 |