TruthX: Alleviating Hallucinations by Editing Large Language Models in Truthful Space

Open paper
Benchmark Model Rank Results
question-answering-on-truthfulqaMistral-7B-Instruct-v0.2 + TruthX#2MC1: 0.56MC2: 0.75
question-answering-on-truthfulqaLLaMa-2-7B-Chat + TruthX#4MC1: 0.54MC2: 0.74

Rank counts only results with a code link.