An adversary can fine-tune an LLM so it behaves normally in full precision but becomes malicious after GGUF quantization, demonstrated across models, bit widths, and attack goals.
Half-quadratic quantization of large machine learning models, 2023
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CR 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Mind the Gap: A Practical Attack on GGUF Quantization
An adversary can fine-tune an LLM so it behaves normally in full precision but becomes malicious after GGUF quantization, demonstrated across models, bit widths, and attack goals.