Empirical comparison finds no single inference-time defense dominates for MLLMs, combinations cause 97-100% over-refusal on benign queries, and adaptive selection based on model and attack type is recommended.
arXiv preprint arXiv:2402.05355 (2024)
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CR 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Comparative Analysis of Inference-Time Defense Methods for Multimodal Large Language Models
Empirical comparison finds no single inference-time defense dominates for MLLMs, combinations cause 97-100% over-refusal on benign queries, and adaptive selection based on model and attack type is recommended.