AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

Unbalanced Optimal Transport for Efficient Visual Document Retrieval

Hoyeon Shin, Jeongyeon Kim, Yeong Jun Koh, Yeoneung Kim, Hanul Kim

ECCV 2026 · Vision-language and multimodal › Multimodal LLMs · approach: Attention-based visual token pruning · 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
replace Light-ColQwen2 hard clustering with UOT compression at MF=50ColQwen2 · ViDoRe V3 cross-lingualaverage nDCG@5: +11.8Improved
replace Light-ColQwen2.5 clustering with UOT compression at MF=50ColQwen2.5 · ViDoRe V3 cross-lingualaverage nDCG@5: +19.0Improved
replace Light-ColQwen2 with UOT compression at MF=50ColQwen2 · ViDoRe V2average nDCG@5: +13.9Improved
replace Light-ColQwen2.5 with UOT compression at MF=50ColQwen2.5 · ViDoRe V2average nDCG@5: +23.7Improved
UOT compression versus pruning at MF=5 (Qwen2.5-VL-3B)ColQwen2.5 · ViDoRe V2average nDCG@5: 49.5 -> 53.1Improved
UOT compression versus pruning at MF=5 (Qwen2-VL-2B)ColQwen2 · ViDoRe V3 cross-lingualaverage nDCG@5: 47.6 -> 48.1Mixed
use hierarchical clustering to compress rather than UOT at MF=25 (V3 Qwen2)Light-ColQwen2 · ViDoRe V3 cross-lingualaverage nDCG@5: 36.2 -> 42.9No clear effect
increase compression factor from 25 to 50UOT method · Indexing timems/doc: 11.4 -> 9.4Improved
increase compression factor from 25 to 50Light-ColQwen2 · Indexing timems/doc: 196.0 -> 200.1Hurt
lower marginal penalty rho toward 0 (ablation)UOT compression · ViDoRe V1 academicretrieval accuracyHurt
relax rho from 100 (balanced OT) to 5 (ablation)UOT compression · ViDoRe V1 academicretrieval accuracyImproved
increase entropy gamma to 0.06 or higher (ablation)UOT compression · ViDoRe V1 academicretrieval accuracyHurt

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