Hidden Tail crafts adversarial images that force VLMs to emit long invisible runs of special tokens, inflating output length up to 19.2x while keeping the visible answer normal.
Early VLMs, such as BLIP [14], InstructBLIP [2], and MiniGPT-4 [36], introduce image encoders (e.g., Vision Transformers) paired with large language models (LLMs) as decoders
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.CR 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
unclear 1representative citing papers
citing papers explorer
-
Hidden Tail: Adversarial Image Causing Stealthy Resource Consumption in Vision-Language Models
Hidden Tail crafts adversarial images that force VLMs to emit long invisible runs of special tokens, inflating output length up to 19.2x while keeping the visible answer normal.