Share your thoughts, 1 month free Claude Pro on usSee more

Truthful and Informative Generation on TruthfulQA (test)

67.14True*Info (%)

Llama3-8B-chat + E/R + Coarse

Updated 1mo ago

Evaluation Results

Method	Links
Llama3-8B-chat + E/R + Coarse 2024.10		67.14
Llama3-8B-chat + FactTune-FS 2024.10		64.58
Llama3-8B-chat + EVER-Pref 2024.10		63.01
Llama3-8B-chat + Self-Eval-SKT 2024.10		61.88
Llama3-8B-chat + SFT 2024.10		59.17
Llama3-8B-chat 2024.10		58.89
Llama2-7B-chat + E/R + Coarse 2024.10		56.47
Llama2-7B-chat + FactTune-FS 2024.10		52.48
Llama2-7B-chat + EVER-Pref 2024.10		51.07
Llama2-7B-chat + Self-Eval-SKT 2024.10		48.65
Llama2-7B-chat + SFT 2024.10		45.52
Llama2-7B-chat 2024.10		38.83