Modelpedia
Work in progress
About
Findings
Models
Concepts
Methods
Datasets
Sources
Light
Dark
Logit Anchoring
anchor
Findings
IC-1206
GPT-2-XL, GPT-J, Falcon-7B, Llama-2-7B, and Llama-2-13B are vulnerable to backdoor injection via lightweight parameter editing with only 15 samples, achieving near-100% attack success rate while preserving clean performance
[compared-to]