List your company

Text Generation on CNN/DailyMail (test)

3.18LCTG Error Rate (E)MARKERGEN
2.09489.419916.74524.0701Feb 19, 2025

Evaluation results

MethodLinks
MARKERGENBackbone=Llama3.1-70B2025.023.183.36—
MARKERGENBackbone=Llama3.1-8B2025.023.363.18—
MARKERGENBackbone=Qwen2.5-32B2025.024.823.25—
MARKERGENBackbone=Qwen2.5-14B2025.026.063.16—
MARKERGENBackbone=Qwen2.5-7B2025.029.923.07—
ImplicitBackbone=Qwen2.5-32B2025.0211.053.21—
ImplicitBackbone=Llama3.1-70B2025.0211.073.09—
ImplicitBackbone=Qwen2.5-14B2025.0212.543.15—
ImplicitBackbone=Llama3.1-8B2025.0215.123.04—
ImplicitBackbone=Qwen2.5-7B2025.0230.313.04—