A 10% curated subset of vision-language instruction data, chosen by a pipeline that unifies image and text scoring through generated captions, matches or slightly beats fine-tuning on the full 500K dataset.
Sharegpt4video: Improving video understanding and generation with better captions.Advances in Neural Information Processing Systems, 37:19472– 19495, 2024
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
-
Better Reasoning with Less Data: Enhancing VLMs Through Unified Modality Scoring
A 10% curated subset of vision-language instruction data, chosen by a pipeline that unifies image and text scoring through generated captions, matches or slightly beats fine-tuning on the full 500K dataset.