Detecting Instruction Fine-tuning Attack on Language Models with Influence Function

TDI
AAML
Re-assign community