Quantizing code LLMs reduces membership inference effectiveness, with 4-bit compression giving larger privacy and performance drops than 8-bit.
Mahoney, and Kurt Keutzer
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.SE 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
How Quantization Impacts Privacy Risk on LLMs for Code?
Quantizing code LLMs reduces membership inference effectiveness, with 4-bit compression giving larger privacy and performance drops than 8-bit.