AUG 18, 2026 · PREPRINT
MemCatalyst: Amplifying Data Auditing on Vision-Language Models via Data Poisoning
arXiv
This is a methods paper proposing a novel data poisoning technique for amplifying membership inference attacks on vision-language models; it presents a computational approach without clinical, patient, or real-world health outcomes.
Reported
Evaluated auditing methodsfive state-of-the-art data audits
VLM architectures testedtwo prominent VLMs
Primary outcomeMI AUC scores marked enhancement