Image autoregressive models leak substantially more training data than diffusion models under membership inference, dataset inference with as few as 4 samples, and data extraction attacks.
IARs benefit from scaling laws (Kaplan et al., 2020), and that allows them to be scaled up to sizes greater than DMs, before their performance plateaus
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2025 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Privacy Attacks on Image AutoRegressive Models
Image autoregressive models leak substantially more training data than diffusion models under membership inference, dataset inference with as few as 4 samples, and data extraction attacks.