DocMemo's tri-level memory with Bayesian page-belief updating and Thompson sampling improves iterative evidence retrieval on three long-document DocVQA benchmarks.
Colpali: Efficient document retrieval with vision language models
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.AI 1years
2026 1verdicts
CONDITIONAL 1roles
background 1polarities
unclear 1representative citing papers
citing papers explorer
-
DocMemo: Dynamic Evidence Discovery via Probabilistic Memory-Guided Retrieval for Multi-Modal Document Understanding
DocMemo's tri-level memory with Bayesian page-belief updating and Thompson sampling improves iterative evidence retrieval on three long-document DocVQA benchmarks.