DOVE adaptively generates a variable number of continuous visual tokens per image via a learned EOS signal, improving efficiency and semantic quality over fixed-length autoencoders.
Pythia: A suite for analyzing large language models across training and scaling
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
-
Images are Worth Variable Length of Representations
DOVE adaptively generates a variable number of continuous visual tokens per image via a learned EOS signal, improving efficiency and semantic quality over fixed-length autoencoders.