RDF-to-text generation on GEM2 Counterfactual (test)
81.8GrammaticalityLlama 3.3
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Llama 3.3Architecture=Neural model2025.12 | 81.8 | 20.9 | 8 | |
| GPT-4.1Architecture=Rule-based NLG2025.12 | 51.7 | 6.9 | 12.8 | |
| Qwen 2.5 72BArchitecture=Rule-based NLG2025.12 | 44 | 5.4 | 9.1 | |
| BARTArchitecture=Neural model2025.12 | 42.6 | 61.3 | 62.2 | |
| Llama 3.3 70BArchitecture=Rule-based NLG2025.12 | 41.9 | 7 | 13.8 | |
| Qwen 3 235BArchitecture=Rule-based NLG2025.12 | 39.2 | 7.1 | 10.6 |