Text Generation on HANNA (test)
2.58LCTG Error RateMARKERGEN
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| MARKERGENBackbone=Llama3.1-70B2025.02 | 2.58 | 3.63 | — | |
| MARKERGENBackbone=Llama3.1-8B2025.02 | 2.98 | 3.6 | — | |
| MARKERGENBackbone=Qwen2.5-32B2025.02 | 3.57 | 3.72 | — | |
| MARKERGENBackbone=Qwen2.5-14B2025.02 | 5.22 | 3.55 | — | |
| MARKERGENBackbone=Qwen2.5-7B2025.02 | 8.49 | 3.5 | — | |
| ImplicitBackbone=Llama3.1-70B2025.02 | 10.44 | 3.61 | — | |
| ImplicitBackbone=Qwen2.5-32B2025.02 | 12.03 | 3.67 | — | |
| ImplicitBackbone=Qwen2.5-14B2025.02 | 14.86 | 3.55 | — | |
| ImplicitBackbone=Llama3.1-8B2025.02 | 16.68 | 3.54 | — | |
| ImplicitBackbone=Qwen2.5-7B2025.02 | 28.55 | 3.47 | — |