Text Generation on CNN/DailyMail (test)
3.18LCTG Error Rate (E)MARKERGEN
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| MARKERGENBackbone=Llama3.1-70B2025.02 | 3.18 | 3.36 | — | |
| MARKERGENBackbone=Llama3.1-8B2025.02 | 3.36 | 3.18 | — | |
| MARKERGENBackbone=Qwen2.5-32B2025.02 | 4.82 | 3.25 | — | |
| MARKERGENBackbone=Qwen2.5-14B2025.02 | 6.06 | 3.16 | — | |
| MARKERGENBackbone=Qwen2.5-7B2025.02 | 9.92 | 3.07 | — | |
| ImplicitBackbone=Qwen2.5-32B2025.02 | 11.05 | 3.21 | — | |
| ImplicitBackbone=Llama3.1-70B2025.02 | 11.07 | 3.09 | — | |
| ImplicitBackbone=Qwen2.5-14B2025.02 | 12.54 | 3.15 | — | |
| ImplicitBackbone=Llama3.1-8B2025.02 | 15.12 | 3.04 | — | |
| ImplicitBackbone=Qwen2.5-7B2025.02 | 30.31 | 3.04 | — |