Modelpedia
Work in progress
About
Findings
Models
Concepts
Methods
Datasets
Sources
Light
Dark
AdversarialGLUE
anchor
Findings
IC-592
The log-likelihood layer in LLaMA-2-7B, LLaMA-2-7B-Chat, Vicuna-7B, and Mistral-7B-Instruct produces factually incorrect answers on TruthfulQA MC1 (817 samples) due to a misalignment between the output distribution and internal attention head representations, with LM-to-head-norm accuracy gaps of 24.23 to 40.68 points.
[eval]