Retrieval-augmented generation improves multi-page document VQA accuracy for small and medium models, with text-based retrieval up to +22.5 ANLS and visual retrieval up to +5.0 ANLS.
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
-
Enhancing Document VQA Models via Retrieval-Augmented Generation
Retrieval-augmented generation improves multi-page document VQA accuracy for small and medium models, with text-based retrieval up to +22.5 ANLS and visual retrieval up to +5.0 ANLS.