AI Paper Atlas

Research map / Vision-language and multimodal / Human and embodied modeling

LightSTAR: Efficient Visual Document Retrieval via Lightweight Selection with Vision-Adaptive Refinement

Tongkun Guan, Haocheng Wang, Wei Shen, Xiaokang Yang

ECCV 2026 · Vision-language and multimodal › Human and embodied modeling · approach: Composed and text-based image retrieval · 1 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 scale-adaptive top-k late interaction in place of max-pooling (ablation)LLM-free Visual Selection · ViDoRe averageRecall@100: 96.8 -> 97.8Improved
add content-grounded query embedding (function-word filtering) (ablation)LLM-free Visual Selection · ViDoRe averageRecall@100: 97.0 -> 97.8Improved
add adaptive region-wise feature fusion with layout-aware features (ablation)Vision-adaptive Semantic Refinement · ViDoRe averageNDCG@5: 88.3 -> 89.1Improved
add hardness-aware objective (ablation)Vision-adaptive Semantic Refinement · ViDoRe averageNDCG@5: 87.9 -> 89.1Improved
add refinement stage to selection-only pipeline (ablation)LightSTAR · ViDoRe averageNDCG@5: 80.8 -> 89.1Improved
replace full-corpus refinement with selection then refinement (ablation)LightSTAR · ViDoRe averageNDCG@5: 89.3 -> 89.1No clear effect
replace full-corpus refinement with selection then refinement (ablation)LightSTAR · ViDoRe latency (5000 pages)latency (s): 465.1 -> 123.9Improved
use unified LightSTAR selection instead of SigLIP first stage before ColQwen2.5cascade pipeline · ViDoRe averageNDCG@5: 81.3 -> 89.1Improved
use unified LightSTAR selection instead of CLIP first stage before ColQwen2.5cascade pipeline · ViDoRe averageNDCG@5: 73.3 -> 89.1Improved
use LightSTAR instead of ColQwen2.5retriever · 7,000-page corpusend-to-end latency: 4.1x faster (656.7/161.1)Improved
use LightSTAR instead of VisRAG-Retretriever · 7,000-page corpusend-to-end latency: 10.4x faster (1678.9/161.1)Improved
use LightSTAR instead of ColPaliretriever · 7,000-page corpusend-to-end latency: 2.3x faster (362.6/161.1)Improved

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