Suffix injection plus PGD image perturbations raises attack success on LLaVA 1.5, with the text suffix contributing most of the gain.
Hallusionbench: You see what you think? or you think what you see? an image-context reasoning benchmark challenging for gpt-4v(ision), llava-1.5, and other multi-modality models
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CR 1years
2024 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Technical Report for ICML 2024 TiFA Workshop MLLM Attack Challenge: Suffix Injection and Projected Gradient Descent Can Easily Fool An MLLM
Suffix injection plus PGD image perturbations raises attack success on LLaVA 1.5, with the text suffix contributing most of the gain.