VLMInferSlow finds small image perturbations that make black-box VLM APIs produce up to 128% longer responses, sharply increasing latency and energy use.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
VLMInferSlow: Evaluating the Efficiency Robustness of Large Vision-Language Models as a Service
VLMInferSlow finds small image perturbations that make black-box VLM APIs produce up to 128% longer responses, sharply increasing latency and energy use.