AI Paper Atlas

Research map / Vision-language and multimodal / Cross-modal learning

CMDR: Contextual Multimodal Document Retrieval

Ryota Tanaka, Taku Hasegawa, Kyosuke Nishida

ECCV 2026 · Vision-language and multimodal › Cross-modal learning · approach: Probabilistic embeddings for cross-modal retrieval · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add contextual joint multi-page encoding and CMCL instead of finetuning only (ablation)ColQwen · CMDR-BenchnDCG@5 overall: 40.4 -> 56.6Improved
add contextual joint multi-page encoding and CMCL instead of finetuning only (ablation)ColPali · CMDR-BenchnDCG@5 overall: 36.3 -> 49.8Improved
add contextual joint encoding and CMCL (ablation)ColQwen · CMDR-BenchRecall@5 overall: 58.8 -> 75.1Improved
finetune backbone on CMDR-Synth (ablation)ColQwen · CMDR-BenchnDCG@5 overall: 33.2 -> 40.4Improved
add CMCL loss with in-chunk and in-document hard negatives (ablation)CMDR-Embed · CMDR-BenchnDCG@5Improved
use late interaction multi-vector instead of single-vector mean pooling (ablation)CMDR-Embed (ColPali) · CMDR-BenchnDCG@5: 19.1-point drop with last-token poolingImproved
add Qwen3-VL reranker to the contextual model (ablation)CMDR-Embed (ColPali) · CMDR-BenchnDCG@5 overall: 49.8 -> 41.0Hurt
add Qwen3-VL reranker to non-contextual model (ablation)ColPali · CMDR-BenchnDCG@5 overall: 32.2 -> 39.5Improved
use GOT-OCR2.0 instead of Tesseract OCR (ablation)BGE text retriever · CMDR-BenchnDCG@5 overall: 29.8 -> 30.0No clear effect
add ViDoRe data via multitask learning (ablation)CMDR-Embed (ColPali) · ViDoRenDCG@5 average: 80.9 -> 85.6Improved
add ViDoRe data via multitask learning (ablation)CMDR-Embed (ColPali) · CMDR-BenchnDCG@5 overall: 49.8 -> 47.1No clear effect
apply hierarchical token pooling factor 5 (ablation)CMDR-Embed (ColPali) · CMDR-Benchvectors / speed / performance: 80.0% fewer vectors, 3.57x faster search, 97.3% performance retainedMixed

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy