A single LLaVA-based captioning model continuously controls caption length, descriptiveness, and word uniqueness by interpolating between learned endpoint conditioning vectors.
nocaps: novel object caption- ing at scale
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.CV 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
unclear 1representative citing papers
citing papers explorer
-
CaptionSmiths: Flexibly Controlling Language Pattern in Image Captioning
A single LLaVA-based captioning model continuously controls caption length, descriptiveness, and word uniqueness by interpolating between learned endpoint conditioning vectors.