List your company

Text Generation on TruthfulQA (test)

2.8LCTG Error RateMARKERGEN
2.1966.27310.3514.427Feb 19, 2025

Evaluation results

MethodLinks
MARKERGENBackbone=Llama3.1-70B2025.022.84.48—
MARKERGENBackbone=Llama3.1-8B2025.023.824.25—
MARKERGENBackbone=Qwen2.5-32B2025.024.484.54—
ImplicitBackbone=Llama3.1-8B2025.027.214.22—
MARKERGENBackbone=Qwen2.5-14B2025.027.594.43—
ImplicitBackbone=Llama3.1-70B2025.027.644.46—
ImplicitBackbone=Qwen2.5-32B2025.028.74.45—
MARKERGENBackbone=Qwen2.5-7B2025.029.084.33—
ImplicitBackbone=Qwen2.5-7B2025.0216.74.29—
ImplicitBackbone=Qwen2.5-14B2025.0217.94.44—