Visual Document Retrieval

Vision-language models (VLMs) are increasingly used to extract structured fields from business documents, yet most evaluations report accuracy on clean benchmarks and offer little guidance to practitioners choosing an approach for a given task complexity. We address this gap with a measurement-grounded study and an open-source release. Across eleven systems (three commercial, two reasoning, five open-source VLMs in pretrained and fine-tuned form, and a non-LLM OCR->regex floor) scored on a 750-document held-out pool of synthetic checks, fine-tuning on 3K samples lifts the best open-source VLMs above F1 0.98-above every zero-shot commercial system on this task-while GPT-5 leads the commercial pool on F1 and Claude Sonnet 4.5 collapses on Date. To turn these measurements into actionable choices, we introduce a practitioner-oriented selection framework that maps a task profile (quality, latency, governance, volume) to a recommended approach via filtering and total-cost minimization, illustrated on a hypothetical mid-volume document-extraction scenario.

Agentic systems can interpret user requests, search the live web, and use external tools, but their ability to transform retrieved web content into a usable 3D world has not been systematically evaluated. No established end-to-end pipeline or benchmark exists for this capability. We introduce Search-to-World, an end-to-end evaluation task covering request understanding, web visual-content retrieval, and 3D-world delivery. We define Observed Retrieval Rate (ORR) and World Delivery Rate (WDR) to distinguish observing relevant content from successfully delivering a request-aligned, perceptually acceptable world. We also present WorldSearcher, a reuse-then-reconstruction harness that connects existing search agents to world delivery: it first retrieves reusable 3D worlds and, when none are available, reconstructs a world from video. A structured recovery controller revises temporal grounding, replaces source videos, or reformulates queries after failure. Using WorldSearcher, we benchmark representative models on Search-to-World and study supervised fine-tuning (SFT) for recovery subagents. Results show that delivery depends on the underlying agentic model, and that relevant-content observation does not ensure world delivery. Jointly training recovery agents improves delivery success and action efficiency. Search-to-World makes agentic 3D-world delivery measurable, while WorldSearcher provides a practical evaluation harness with recovery capabilities.

Late-interaction visual document retrievers preserve fine-grained page evidence by storing many token embeddings per page, but the resulting storage and query-time interaction costs make large-scale deployment expensive. Pooling document tokens before indexing offers a natural remedy, yet static pooling must decide which visual evidence to preserve before the query is known. We study an alternative: a heavily compressed hot-path index generates candidates, after which query-aware token budgeting operates on the original token sets of the shortlisted pages. We formulate this stage-two selection as a budgeted MaxSim coverage problem, show that a clipped version is monotone submodular, and compare coverage-only, cluster-guided, token-wise, and marginal-gain policies. On ten ViDoRe tasks with ColModernVBERT, direct static pooling reduces macro normalized discounted cumulative gain at rank five from 0.6309 without compression to 0.4738 at a thirty-two-fold pool factor. Under the same candidate-generation regime and a pool-factor-eight-equivalent reranking budget, token top-k recovers 93.93 percent of the full-token score, while greedy marginal-gain selection recovers 98.39 percent. Held-out and leave-one-dataset-out evaluations yield positive greedy improvements over token top-k on every dataset. The latency analysis reveals two useful operating points: token top-k for interactive retrieval and the naive greedy implementation as a quality upper envelope. Together, these results show that late-interaction visual retrieval benefits from query-aware allocation rather than query-agnostic pooling alone.

Multi-modal late-interaction retrievers achieve strong retrieval on visually rich documents by representing each page as per patch embeddings and matching at the token level. However, this approach incurs high storage costs. Existing compression methods typically fix a single compression level at indexing time, limiting flexibility. We present ColSNAP (Spatial Nested Average Pooling)1, a training method that generates a nested hierarchy of compression levels directly from a backbone's patch grid. By spatially pooling patch embeddings into pro- gressively coarser tiers and training all tiers simultaneously, a single model learns to support retrieval at multiple compression levels without architectural changes. Crucially, a single encoding pass yields every tier, enabling the accuracy-storage trade-off to be configured at indexing time to match avail- able storage budgets, rather than being fixed during training. We demonstrate that models trained using ColSNAP maintain near full-resolution retrieval performance under substantial compression and that ColSNAP transfers effectively across multiple late-interaction backbones, and achieves most of its improvements via a lightweight adaptation stage applied to a pre-trained retriever.

While the privacy risks of multimodal large language models (MLLMs) have drawn significant attention, the unique vulnerabilities of domain-specific MLLMs remain largely underexplored. Focusing on document understanding MLLMs for identity document processing, this paper investigates the privacy issues inherent in Key Information Extraction (KIE) tasks. We reveal that when input images lack sufficient visual evidence, these models often rely on memorized field relations from training data to infer missing content, thereby leaking multiple correlated fields containing sensitive personal information. To mitigate this risk, we make three key contributions.First, we propose the Dynamic Relational Unlearning Framework (DRUF) which comprises a Relational Decoupling Unlearning (RDU) module and a dynamic set update mechanism. It suppresses the leakage of high-risk field pairs while preserving KIE performance.Second, we introduce DocPrivacyBench, a novel benchmark to systematically evaluate a model's susceptibility to privacy leakage under conditions of absent or minimal visual evidence.Third, we evaluate three MLLMs and six unlearning methods using this benchmark, assessing both post-unlearning leakage suppression and utility preservation.Our results demonstrate that existing MLLMs consistently exhibit privacy leakage when visual evidence is scarce, particularly on noisier datasets. In contrast, DRUF outperforms the strongest baseline by improving leakage suppression by 4.8 percentage points, effectively mitigating privacy risks while maintaining robust document information extraction performance.

Visual document retrieval (VDR) is dominated by multi-billion-parameter models that are slow to index at full corpus scale and expensive to serve. Prior compression routes either train a smaller multi-vector encoder from scratch or distil only the query side; neither yields a compact single-vector retriever end-to-end. We present DistilVDR, a 524M end-to-end VDR system distilled bilaterally from a single 8B vision-language teacher under a pointwise cosine alignment loss. All supervision comes from the frozen teacher's embedding space, which was itself trained with relevance supervision, so the student objective needs no relevance labels, negative sampling, or contrastive term. We match VDR's text-query and image-document input asymmetry with an asymmetric encoder-only student that concentrates visual capacity on the document side and keeps the query side at 70M parameters. We release two variants that share the same encoders and training and differ only in the document encoder's visual-tile budget: DistilVDR-HiRes attains 61.74 average NDCG@5 on ViDoRe v1+v2+v3 (86.9% of the 8B teacher) and leads every reproduced sub-1B baseline on the high-resolution-sensitive v3 benchmark, while DistilVDR-Fast attains 59.98 at a 3 times smaller visual-token budget. Both variants store one million documents in a 15.6 times smaller index than the strongest sub-1B multi-vector baseline and index the corpus an order of magnitude faster. The code is available at https://github.com/Ryenhails/NanoVDR.

Long-document understanding often requires reasoning over many visually rich pages, making inference costly and prone to context rot. In this work, we propose InSight-doc, an agentic visual perception framework that treats visual resolution as an adaptive reasoning-time resource. InSight-doc starts from low resolution and selectively zooms into high-resolution regions for finer evidence, without relying on any external retriever. To train such an agent, we construct an active-perception corpus of 17.9K high-quality SFT examples with region-level zoom-in trajectories, accompanied by 19.2K hard RL examples. Through SFT+RL, InSight-doc-8B improves the baseline by 4.3--16.4 accuracy points over document VQA benchmarks. On long documents, it reduces hallucination by more than 40% and inference latency by 41%--68% while maintaining an accuracy lead. Our code, datasets, and model are released at https://github.com/m-Just/InSight-doc .

Multi-vector vision-language retrievers enable fine-grained Visual Document Retrieval (VDR) through late interaction, but storing and scoring hundreds of visual patch embeddings per page incurs substantial overhead. Existing training-free methods rely on pruning or merging: pruning degrades sharply under aggressive compression, whereas merging does not explicitly prioritize important regions when forming representatives. We introduce AnchorFold, a training-free focus-then-fold framework for document-side index compression. AnchorFold applies Recursive Attention Propagation over visual self-attention graphs, performing multi-step propagation within each attention head and integrating scores across heads and layers. The focus stage selects the highest-centrality tokens as anchors. The fold stage assigns remaining tokens to their most similar anchors in the normalized retrieval space and summarizes each anchor-centered group through centrality-weighted aggregation. This preserves non-anchor contributions while concentrating capacity on structurally important tokens. Across ViDoRe v1/v2 and REAL-MM-RAG with three diverse retrieval backbones, AnchorFold consistently outperforms all evaluated training-free baselines at $γ\leq 0.20$. On ViDoRe v1/v2, it retains 98.3% of full-index NDCG@5 on average at $5\times$ compression, achieving near-lossless compression, and 92.4% at $20\times$ compression.

Multi-page visually-rich document understanding (MP-VRDU) requires managing evidence that is sparse, spread across pages, and often exceeds a model's context window. Prior work has produced competing, largely untested claims about how these systems should be built. We attribute incorrect answers to three failure modes, representation, selection, and reasoning, and isolate each over a multi-page document understanding dataset by intervening on one while holding the others fixed. We find that vision is necessary but does not replace text extraction, that missing pages bound accuracy while distractors cost little, and that reasoners fail to integrate evidence across pages even when it is fully supplied. Prompting can shift reasoning behaviour substantially, improving some outcomes at the expense of others. We translate these findings into guidance for building such systems under a fixed compute budget.

Contrastive vision-language models such as CLIP and BLIP are typically trained on short image captions, limiting their ability to retrieve images from detailed textual descriptions. While methods such as Long-CLIP extend the token limit through positional embedding interpolation, we ask a simpler question: does training text granularity alone determine long-text retrieval performance? We present a systematic study of supervision ranging from single captions to multi-sentence paragraphs for contrastive image-text retrieval. Using a synthetic pipeline based on Qwen2-VL and Llama 3.2 Vision, we generate diverse captions, hard negatives, and quality-scored paragraphs for 500K CC3M images. To isolate the effect of text granularity, we fine-tune only the BLIP text encoder while keeping the vision encoder frozen across 10 training configurations. Our paragraph-supervised models match Long-CLIP-L on ShareGPT4V and outperform it by more than 14 points on DOCCI for image-to-text retrieval, without architectural changes. We further show that paragraph supervision enables effective use of long token sequences, whereas caption-only training degrades beyond 60 tokens. Increasing caption diversity improves short-caption retrieval with diminishing returns, while paragraph supervision consistently benefits long-description benchmarks and hard negatives prove detrimental in text-only fine-tuning. Evaluations on Flickr30k, COCO, ShareGPT4V, and DOCCI provide a comprehensive analysis of the trade-offs between text granularity, retrieval direction, and description length.

Existing document understanding benchmarks have largely focused on locating page elements, yet real-world document intelligence requires models to reason jointly about region semantics, spatial relations, and visual structure. We present ADOPD 2026, a reasoning-oriented extension of ADOPD that turns page decomposition into spatially grounded document understanding. ADOPD 2026 enriches page anchors inherited from ADOPD 2024 dataset with human-cleaned captions, semantic tags, and generated chain-of-thought (CoT) traces grounded to document regions. Instead of treating boxes, masks, and tags as independent supervision signals, we cast text blocks, visual entities, semantic labels, bounding boxes, and polygon masks as a shared vocabulary of visual anchors. This representation supports three connected capabilities. First, region-level semantic tagging asks models to identify document element types from both page context and local appearance, revealing long-tail semantic failures that standard layout benchmarks often hide. Second, unified vision-language grounding generates text regions and visual entities together with coordinates or polygonal outlines, transforming detection and segmentation outputs into structured anchors that can be reused by downstream reasoning systems. Third, current state-of-the-art models still struggle with dense counting tasks evaluated on DocCount, a benchmark derived from ADOPD 2026, highlighting the need for the Thinking-with-Anchors pipeline in document semantic understanding. By connecting page decomposition to verifiable visual-anchor reasoning, ADOPD 2026 provides a task framework that moves document understanding beyond localization toward anchor-grounded document intelligence.

Visual Document Retrieval (VDR) directly matches text queries against document images, preserving visual and structural information that may be lost during text extraction. However, existing VDR models and training resources remain predominantly English-centric, while many high-performing systems rely on massive backbones or storage-intensive multi-vector representations. To address these limitations, we introduce KoVRE: Korean Visual Document Retrieval Embedding, a single-vector retriever for Korean visual documents, alongside a comprehensive training recipe. We train the model on 708,729 Korean and English query-page pairs using positive-aware hard-negative mining and conduct controlled analyses of training-data composition, hard-negative treatment, and reranker-based knowledge distillation. Across Korean visual document retrieval benchmarks, our 2B model substantially improves over the base backbone model, outperforming both its 8B single-vector counterpart and a strong multi-vector baseline. These results demonstrate that targeted bilingual supervision and our carefully designed training strategies can produce a highly effective Korean VDR model across diverse document domains, without requiring a scaled-up backbone or multi-vector representations.

Visual document retrieval has recently become increasingly important in applications such as enterprise search, scientific literature discovery, and retrieval-augmented generation. These applications depend on efficiently identifying query-relevant pages across large collections of visually rich documents. Existing methods commonly adopt late-interaction architectures that encode and index documents offline to enable scalable and low-latency online retrieval. Despite its efficiency, this paradigm requires each document to be encoded into a fixed representation before the query is known. However, the same content in a visual document may induce different interpretations depending on the query intent, which a fixed representation struggles to capture. Yet postponing document encoding until the query arrives would incur prohibitive online retrieval latency. To address this gap, we propose VaRS-Doc, a visual document retrieval framework that diversifies document representations by enabling the model to actively explore variant latent interpretations during document encoding, while preserving efficient late-interaction retrieval in which each query adaptively selects the best-fit representation. We further introduce a two-stage training strategy that encourages the model to capture complementary semantic interpretations and prevents it from falling back to train a single dominant representation. Experiments on visual document retrieval benchmarks show that VaRS-Doc achieves state-of-the-art retrieval performance, offering a practical solution to the mismatch between query-agnostic document encoding and query-specific retrieval needs. Code is available at https://github.com/bokufa/VaRS-Doc.

Multi-page document visual question answering requires locating sparse evidence at both the page and region levels. Existing approaches typically emphasize one level over the other: page-centric methods focus on page acquisition, with region operations serving mainly as navigation aids, whereas region-centric methods assume that the relevant pages have already been supplied. Consequently, page and region selection remain disconnected rather than forming successive evidence decisions. We propose HierDoc, a hierarchical evidence-routing framework that formulates long-document evidence acquisition as two-stage set prediction from pages to regions. A page policy selects evidence pages from the full document; these pages are then parsed for semantic elements, after which a region policy selects the elements passed to a downstream answer model. Both answer-agnostic policies are optimized with stage-wise GRPO using granularity-specific structured-set rewards. The answer model receives selected full pages together with selected region crops and OCR or table text, preserving global context while emphasizing fine-grained evidence. Across the evaluated benchmarks, HierDoc achieves state-of-the-art or competitive performance among open-weight systems, improving LongDocURL by 16.87% relative to the strongest reported open-weight baseline. Controlled ablations further show that selected regional evidence improves the page-only system in accuracy and F1 by 5.51% and 4.82%, respectively. These results demonstrate the benefit of organizing coarse page routing and fine-grained region routing as successive, separately optimized stages of a unified evidence-acquisition process.

In demanding professional environments and meeting review scenarios, lengthy text often imposes a high cognitive load. To facilitate efficient information communication, transforming verbose text into logically clear diagrams is essential. Scalable Vector Graphics (SVG) provide an effective representation for this purpose due to their editability and resolution independence. However, current research on Text-to-SVG generation remains hindered by three major challenges: (1) the scarcity of datasets for complex, logic-rich diagrams; (2) the absence of explicit layout priors, which leads to chaotic spatial arrangements; and (3) the lack of fine-grained visual feedback to validate rendered outputs and correct aesthetic defects. To address these challenges, at the data level, we introduce DocMeetSVG-100K, a large-scale SVG dataset tailored for document authoring and meeting review scenarios. At the model level, we propose GVR-Coder, a novel framework designed to generate high-quality logical diagrams from lengthy professional texts. Specifically, we adopt a curriculum-driven rejection sampling fine-tuning to progressively enhance the model's capability in modeling complex structures, while explicitly incorporating layout constraint knowledge during training. In addition, we introduce reinforcement learning from dual rendering feedback, a mechanism that provides implicit feedback through reward signals to jointly optimize structural complexity and visual aesthetics. Furthermore, we design a generate-verify-repair agent loop, which improves generation quality through explicit, fine-grained feedback and targeted refinement. Extensive experiments demonstrate that GVR-Coder outperforms competitive baselines and reliably produces logically coherent and visually appealing diagrams. Code and data are available at https://github.com/CurryaNa/GVR-Coder.

Reliable visual document understanding requires a model to attribute each answer to the evidence regions that support it. Recent benchmarks and systems express this step through a coordinate interface: the model outputs the coordinates of bounding boxes that mark the evidence regions in the document. Under this interface, vision-language models often fail to identify the right regions even when the answer is correct, a failure known as Attribution Hallucination. We present a study that investigates whether this failure is partially limited by what the model can express through coordinates. On a verified bilingual CiteVQA subset, we compare the coordinate interface with a language interface in which the model outputs only text, quoting its evidence verbatim, and a multimodal retriever returns the location of each quote as a page region proposed by a layout parser (tables and figures are quoted through their captions or notes); the comparison is repeated over six open vision-language models. Compared with the coordinate interface, evidence recall rises from at most 8 points to between 26 and 47 and the hallucination rate roughly halves, with little change in answer quality. Building on this comparison, we use the same quote-and-retrieve pipeline as a training scaffold: because region-level evidence labels are expensive to collect for long documents, we introduce a GRPO recipe whose reward is a judge's reading of the gold answer and crops of the retrieved regions, training the model to quote better evidence without any region labels and raising an 8B backbone's strict attributed accuracy from 22.4 to 33.8. These findings indicate a practical path to improve attribution"without a coordinate interface and without costly region-level supervision.

Visual information extraction (VIE) from visually rich documents remains challenging due to high layout variability and real-world impairments. Existing methods typically rely on sequential OCR pipelines or end-to-end models requiring extensive labeled data and layout-specific training, limiting their scalability.We propose a classification-guided large vision-language model (LVLM) framework for multi-type VIE that achieves high accuracy with minimal supervision. The approach decouples document-type classification from content extraction and employs in-context learning (ICL)-based dynamic prompt engineering to inject task-specific knowledge, enabling robust zero-shot inference across diverse layouts. From a theoretical perspective, the proposed method can be viewed as a form of conditional computation that reduces task uncertainty and improves information efficiency during prompt-based inference. Evaluated on a real-world bidding dataset with 16 certificate types, our zero-shot method (based on Qwen2.5-VL-7B) outperforms a strong supervised baseline by 18.35 percentage points in F1-score (86.43% vs. 68.08%) and 0.23 in normalized edit distance (0.90 vs. 0.67). Optional domain-specific fine-tuning further improves performance to 93.65% F1 and 0.93 NED, demonstrating superior robustness against seals, watermarks, and low contrast. The framework offers an efficient, scalable solution for complex document understanding in office automation. Code is available at https://github.com/FairmeHIT/Multi-VIE, and fine-tuned models at https://huggingface.co/fairme/Qwen2.5-VL-7B-SFT.

Vision language models (VLMs) have achieved strong performance on visual document understanding benchmarks such as DocVQA, ChartQA, and MMLongBench-Doc. However, real-world documents combine multiple factors such as length, layout complexity, modality, and question difficulty, which makes it difficult to attribute model failures to specific causes. We introduce SynthDocBench, a fully synthetic benchmark for long-context visual document understanding that systematically controls factors including document length, layout structure, modality composition, and question type. The benchmark is constructed using a combinatorial design, each factor is varied independently across generated documents, enabling controlled analysis of model behavior. Documents are generated end to end using an LLM pipeline across six layout archetypes, with a 40 percent random override to prevent models from exploiting spurious correlations. Additionally, SynthDocBench spans long-context documents with substantially greater length and structural diversity than existing benchmarks. Evaluating seven frontier VLMs, we uncover three failure modes that existing benchmarks cannot surface: sharp degradation with document length, a systematic positional sensitivity in which the middle third of a document is hardest for five of six models and five of six models show a negative Early-to-Late trend (steepest decline: 8.3 percentage points), and breakdown of chart comprehension in long-document settings. These results suggest that current models may be overfitting to benchmark artifacts rather than achieving robust long-context visual document understanding.

Retrieval-Augmented Generation (RAG) streamlines long-document understanding by leveraging retrieval mechanisms to restrict input images to a highly curated subset. However, existing multimodal RAG pipelines primarily face two critical challenges: first, standard semantic similarity retrievers frequently fetch topically overlapping yet answer-void distractor pages that mislead downstream generation; second, rigid single-pass pipelines heavily depend on initial retrieval success, where any omission of core evidence inevitably causes cascading errors. To address these challenges, we introduce HIEVI-RAG, a hierarchical, evidence-driven multimodal RAG framework for closed-domain document understanding. HIEVI-RAG systematically factorizes complex queries into a cooperative four-stage pipeline: (1) hierarchical question decomposition to break multi-hop root queries into atomic child questions; (2) coarse visual page retrieval leveraging a multimodal retriever to fetch candidate pages based on semantic similarity; (3) fine-grained page verification via EVIAGENT, a specialized multi-page verifier trained with GRPO to execute cross-page reasoning over multi-image blocks; and (4) memory-guided iterative generation that leverages accumulated sub-question context to execute multi-round, dynamic reasoning over the prioritized sequence. Extensive evaluations across four benchmarks demonstrate the robust efficacy and synergy of our framework, which significantly outperforms existing open-source baselines and exceeds the strongest reported baseline by an average of 8.05% in accuracy.

Multimodal large language models (MLLMs) are widely applied to visual document understanding. However, comprehending long documents remains an issue by the limited context window. Though recent multimodal retrieval-augmented generation (MMRAG) can address this challenge by retrieving relevant pages. It still struggles with the visual question answering (VQA) requiring holistic comprehension of a document. To cope with this, knowledge graph (KG) that summarizes global knowledge of a document can provide an effective solution. However, most existing LLM-based KG construction methods handle only the language modality, leaving the automatic creation of multimodal KGs (MMKGs) for visually rich documents largely unexplored. In this paper, we introduce a multimodal graph-based RAG approach to tackle this problem. Existing LLM-based KG methods evaluate the QA performance relying on indirect evidence such as comprehensiveness, diversity, empowerment, and so on. The lack of annotated datasets for comprehensive document-level VQA poses a significant challenge to effective model evaluation. To overcome this limitation, we also introduce a new benchmark, DLVQA (document-level VQA), which provides reference summaries and corresponding supporting facts for global document-level questions. Experimental results show that our approach outperforms existing MMRAG or KG-based approaches on multi-hop QA/VQA benchmarks and DLVQA.

Vision Language Models have achieved near-human performance on single-document Visual Question Answering, yet their effectiveness degrades significantly when retrieving information from large collections of visually homogeneous documents. Existing multi-document benchmarks aggregate diverse document types, creating artificial separation in embedding space that does not reflect enterprise document repositories where thousands of records share identical visual templates. We identify this as embedding collapse and introduce Invoice Haystack, a benchmark with 1,500 anonymized invoice images paired with 200 discriminative question-answer pairs, specifically designed to stress-test retrieval under strong visual homogeneity. Invoice Haystack exhibits a mean pairwise cosine similarity of 0.73, compared to 0.38 (DocHaystack) and 0.31 (InfoHaystack) in existing benchmarks, posing a fundamentally more challenging retrieval problem. Addressing the identified challenge, we propose VL-RAG, a hybrid retrieval-augmented generation framework that jointly leverages text and visual embeddings to harness the complementary strengths of both modalities, followed by a VLM-based verification filter for precise document identification. VL-RAG achieves 60.0% Recall@1 on Invoice Haystack-500, outperforming existing state-of-the-art method by up to an absolute 13.5 percentage points. It further improves retrieval considerably on DocHaystack-1000 (77.1% vs.\ 75.2%) and InfoHaystack-1000 (84.5% vs.\ 80.0%), establishing the proposed dual-stream fusion as a consistently superior retrieval strategy across both homogeneous and heterogeneous document collections.

Visual grounding in documents is a crucial ability for Large Multimodal Models (LMMs) in areas such as document understanding, deep research and document error detection. However, existing approaches exhibit poor grounding precision in text-rich document images, often failing to accurately locate the critical document elements needed for reliable reasoning. To address this gap, we introduce PreciseDoc, an LMM specifically designed for precise element grounding and can be further optimized for Document VQA tasks. Specifically, to enhance the basic localization capability, we construct challenging training data by two pipelines capable of mass-producing high-quality documents with paired metadata of fine-grained coordinates, including synthetic hand-filled documents with camera effects. The model develops more real-world functions beyond straightforward localization of single text, such as locating personal information from CVs. Furthermore, we introduce a training paradigm for visual grounded reasoning where the grounding and reasoning are supervised jointly with reinforcement learning to improve the contribution of the grounded evidence. A comprehensive evaluation on various benchmarks demonstrates the advantage of the proposed data and methods in document spatial grounding and document understanding.

Visual document retrieval requires rapidly locating relevant pages from large multi-modal corpora in response to user queries. While recent methods powered by Multi-modal Large Language Models (MLLMs) show competitive accuracy, they suffer from prohibitive computational costs by applying intensive MLLM encoding to every single page. Meanwhile, we observe that user queries are typically keyword-anchored, containing semantically rich words that are expected to appear directly in the visible text of relevant pages, offering an efficient cue for quickly narrowing down candidate pages. Building on this insight, we propose LightSTAR, an efficient framework that decomposes visual document retrieval into: 1) LLM-free Visual Selection, which utilizes content-grounded query encoding to focus on informative words and employs LLM-free visual embeddings to produce a high-recall candidate set; and 2) Vision-adaptive Semantic Refinement, which further performs fine-grained semantic matching exclusively on these top candidates via adaptive region-wise feature fusion to effectively combine textual and layout cues, optimized through a hardness-aware contrastive objective. Experimental results demonstrate that LightSTAR achieves state-of-the-art retrieval accuracy while reducing end-to-end latency by several-fold, offering a highly practical solution to the accuracy-efficiency trade-off in visual document retrieval. Code is available at https://github.com/bokufa/LightSTAR.

In this technical report, we focus on solving the challenge of Vietnamese multi-document abstractive summarization, introduced in the International Workshop on Vietnamese Language and Speech Processing (VLSP) 2022. We choose to follow the popular hierarchical approach, i.e. condensing each document followed by aggregation and summarization. We propose a novel yet simple strategy to shorten documents that is driven by the golden summary, thus ensuring high correlation between stages of the hierarchical approach. Our method achieves a ROUGE2-F1 score of 0.2468 on the VLSP's public test set, and can produce fluent and concise summaries. Additionally, we utilize external sources for extra data, which greatly enhances the quantity of data for Vietnamese multi-document summarization. The additional data is made available for the community.

Multi-vector visual document retrievers achieve strong fine-grained matching by representing each page with multiple vectors from deep Vision-Language Models (VLMs), but this design makes deployment expensive in both storage and computational overhead. Existing efficiency techniques usually optimize only part of this budget, leaving multimodal retrievers without a unified way to trade accuracy for both vector width and encoder depth. Therefore, we propose MM-Matryoshka, a 2D Matryoshka training framework for budget-elastic Visual Document Retrieval (VDR), enabling ColPali-style multi-vector retrieval elastic along both dimension and layer. At inference time, a single retriever can select a 2D selectable budget without training separate models for different budgets. Through comprehensive experiments across multiple representative backbones, we demonstrate that by retaining significantly higher quality than direct truncation baselines while substantially reducing storage and computational overhead, MM-Matryoshka can offer robust budget elasticity for efficient VDR.

Retrieval over visually-rich documents, pages that interleave text with figures, tables, and charts, is essential for multimodal retrieval-augmented generation, yet most retrievers still discard the visual channel. The \emph{Multimodal Document Retrieval Challenge}, Track~1 of the MIR Challenge at the first EReL@MIR workshop, co-located with The Web Conference 2025, asks participants to build a \emph{single} retrieval system that handles two complementary regimes: closed-set document page retrieval within long documents from a text query (MMDocIR), and open-domain retrieval of Wikipedia-style passages from an image or image-plus-text query (M2KR). Systems are ranked by the macro-average of mean Recall@$\{1,3,5\}$ over the two tasks. The challenge drew 455 entrants and 586 submissions across 22 teams. This report describes the challenge design, datasets, and evaluation protocol; reports the final standings; and analyses the three winning teams' systems. All three build on decoder-based Multimodal-LLM embedders from the Qwen2-VL family rather than on CLIP-style encoders, and differ chiefly in whether they reach the top through fine-tuned ensembles, training-free multi-route fusion with a strong vision-language re-ranker, or zero-shot late interaction. The training-free system finished within $0.1$ point of the fine-tuned winner.

Augmenting large language models (LLMs) with retrieved web text has become a dominant paradigm, yet the web is not natively textual: existing systems depend on complex parsing pipelines that linearize HTML and discard layout, visual structure, and formatting. We introduce PixelRAG, a new retrieval-augmented method that represents websites in their native visual form and performs retrieval and reading entirely in pixel space, enabling an end-to-end architecture that eliminates text abstraction. PixelRAG is, to our knowledge, the first pipeline to operate over a full Wikipedia corpus in this form, scaling to a datastore of 30 million screenshot images with an efficient visual retrieval index. Built on an existing visual embedding model (i.e., Qwen3-VL-Embedding), PixelRAG further fine-tunes this model on screenshot data with carefully curated contrastive training data. Retrieved screenshots are then fed directly as pixel inputs to a VLM, without intermediate text conversion. PixelRAG consistently outperforms both no-retrieval and text-based RAG baselines, most surprisingly on widely studied text-centric tasks such as NQ and SimpleQA. It also achieves strong gains on multimodal open-domain QA (e.g., MMSearch), benchmarks over noisy news corpora (e.g., LiveVQA), and agentic benchmarks (e.g., MoNaCo), improving accuracy by up to 18.1% over text-based baselines. Finally, pixel representations enable a new efficiency lever for RAG through image compression, achieving up to 3x token cost reduction at lower resolutions while maintaining accuracy. Our results challenge the necessity of text representations in web retrieval, suggesting that web RAG can operate directly in the web's native visual form while improving both performance and efficiency.

Document type classification in visually rich documents remains challenging, as relevant information is distributed across textual, visual, and layout modalities. To capture this complexity, current approaches rely on diverse multimodal modeling strategies, resulting in heterogeneous architectures that complicate systematic comparison. This variability is also reflected in existing comparative studies, which often rely on heterogeneous evaluation setups, further complicating systematic comparison and making it difficult to assess progress. To address these limitations, this work provides a structured analysis of multimodal design strategies across transformer- and LLM-based architectures, combined with a controlled empirical comparison within a unified experimental framework. Specifically, four representative models (LayoutLMv3, Donut, Qwen3-VL-32B-Instruct, and Qwen3-32B) are evaluated on the RVL-CDIP benchmark to systematically analyze the contributions of text, image, and layout information for document type classification, with a particular focus on contrasting OCR-dependent and OCR-free approaches. The results show that specialized multimodal Transformers outperform LLM-based approaches on visually rich and layout-intensive documents. Image information contributes most strongly to reliable classification, while OCR-derived text provides useful but secondary support. These findings highlight that multimodal processing remains essential for documents with pronounced layout structure. Overall, the study provides a systematic basis for comparing multimodal architectures and offers practical guidance for selecting effective feature combinations and model designs for document type classification.

As large-scale visual-document corpora such as arXiv papers and enterprise PDFs continue to grow, visual-document retrieval has gained increasing attention; yet it still lacks a deployable system that lexically indexes visual documents to serve queries without neural encoding at scale. Existing methods either achieve strong retrieval quality with VLM-based dense or multi-vector models but require neural query encoding at serving time, or avoid query encoding with OCR- or caption-based BM25 at the cost of time-consuming text extraction or generation. To fill this missing serving regime, we present V-SPLADE, an inference-free sparse retriever for visual-document retrieval. However, such inference-free multimodal learned sparse retrieval systems remain underexplored and have not yet shown dense-level effectiveness under high sparsity. We attribute this limitation to a lexical grounding problem: visual sparse representations often fail to capture the lexical content embedded in document images. To address this problem, we introduce caption-gated token supervision, a training-only signal that uses VLM-generated captions as lexical cues to activate retrieval-relevant vocabulary dimensions. With this supervision, V-SPLADE improves average NDCG@5 across six visual-document retrieval benchmarks by +13.8pp over the same-scale dense baseline and by up to +6.3pp over OCR- or caption-based BM25 baselines. On an 18.7M-document corpus, it more than doubles R@5 over the same-scale dense baseline and further improves competing retrievers through score fusion by up to +2.4pp R@5. Code will be released soon at https://github.com/naver/v-splade.

Document retrieval in real-world scenarios faces significant challenges due to diverse document formats and modalities. Traditional text-based approaches rely on tailored parsing techniques that disregard layout information and are prone to errors, while recent parsing-free visual methods often struggle to capture fine-grained textual semantics in text-rich scenarios. To address these limitations, we propose \textbf{Unveil}, a novel visual-textual embedding framework that effectively integrates textual and visual features for robust document representation. Through knowledge distillation, we transfer the semantic understanding capabilities from the visual-textual embedding model to a purely visual model, enabling efficient parsing-free retrieval while preserving semantic fidelity. Experimental results demonstrate that our visual-textual embedding method surpasses existing approaches, while knowledge distillation successfully bridges the performance gap between visual-textual and visual-only methods, improving both retrieval accuracy and efficiency.

Extracting structured information from visual documents (Visual Information Extraction, VIE) is a cornerstone of business automation. While recent Multimodal Large Language Models (MLLMs) have shown promising capabilities, existing benchmarks suffer from critical limitations in scale and realism, lack semantic granularity, and fail to cover diverse document types. To bridge this gap, we introduce ReceiptBench, a large-scale, human-annotated benchmark consisting of 10k diverse receipts, organizing information extraction into four hierarchical sub-tasks: (1) Basic Perception for raw text spotting, (2) Format Normalization for strictly following standardization instructions, (3) Semantic Reasoning for inferring implicit attributes from context, and (4) Structure Parsing for handling nested line items. Furthermore, we propose a two-stage training framework incorporating Metric-Aware Group Relative Policy Optimization (GRPO), which translates rigorous evaluation constraints into reinforcement learning signals to enhance structural consistency. Extensive experiments demonstrate that our method yields state-of-the-art performance, surpassing leading proprietary models on complex reasoning tasks. We release our datasets and code at https://github.com/wwwT0ri/ReceiptBench.

Vision-Language Models (VLMs) parse documents end-to-end but frequently break down on layouts unlike those seen in training. We attribute this to a two-hop bottleneck: before the decoder can extract content (Hop 2), it must first classify and localize the enclosing layout entity (Hop 1), and when the first hop fails the second collapses into omissions, malformed structure, or autoregressive repetition. We pre-resolve Hop 1 outside the decoder by running a lightweight RT-DETR detector, serializing its outputs in the parser's native DocTags vocabulary, and injecting them into the prompt alongside the full page image. Unlike analyze-then-parse approaches that crop the page, or prior prompt-level priors written in plain text, our prior shares the decoder's generation space and leaves the global image in view as a fallback when detections are noisy. On a 10k-page structural out-of-distribution benchmark, markdown F1 rises from $0.37$ to $0.92$; on the Chinese subset of OmniDocBench, table TEDS rises from $0.01$ to $0.36$; and on the 26k-page ViDoRe V3 benchmark, infinite-loop decoding failures drop across every industrial domain tested. These gains cost $15\%$ wall-clock latency and a median of $74$ prompt tokens, with no architectural change to the base VLM. An attention-level analysis further reveals a bimodal phase shift in which the decoder attends to injected layout tokens when emitting structure and to image patches when emitting content, consistent with the two-hop bottleneck being alleviated. Model weights will be released to support reproducibility.

Vision-Language Models (VLMs) have shown strong promise on Optical Character Recognition (OCR), yet the sheer number of visual tokens required to encode dense documents incurs prohibitive inference cost. Existing pruning methods rely on physical eviction, e.g., permanently discarding visual tokens during the prefill stage. While effective for natural images, this strategy fundamentally breaks down on OCR, where virtually every visual token may correspond to a character or structural element, and any irreversible loss leads to catastrophic accuracy degradation. We observe that, although document images appear globally dense and seemingly unprunable, the model's attention to them is in fact temporally sparse: at each decoding step it concentrates on a small region that shifts gradually across steps, much as a human reader fixates on successive words rather than perceiving an entire page at once. Motivated by this Dynamic Visual Fixation phenomenon, we recast the intractable global pruning problem as a tractable local, dynamic one and propose FastOCR, a training-free framework with two complementary modules. Specifically, Focal-Guided Pruning identifies a small set of focal layers and selects the most task-relevant visual tokens from them at each step, while Cross-Step Fixation Reuse exploits the gradual shift of fixation to warm-start each step from the previous one. By dynamically adjusting which tokens are attended rather than evicting any from the cache, FastOCR avoids permanent information loss. Extensive experiments show that FastOCR serves as a plug-and-play acceleration module, generalizing consistently across five VLMs of varying sizes and architectures. On Qwen2.5-VL, FastOCR retains 98% of the unpruned model's accuracy while attending to only 5% of the visual tokens per decoding step, reducing attention latency by 3.0$\times$.

Visual RAG has offered an alternative to traditional RAG. It treats documents as images and uses vision encoders to obtain vision patch tokens. However, hundreds of patch tokens per document create retrieval and storage challenges in a vector database. Practical deployment requires aggregating them into a single vector. This raises a critical question: does single-vector aggregation lose key information in financial documents? We develop a diagnostic benchmark using financial documents where changes in single digits can lead to significant semantic shifts. Our experiments show that single-vector aggregation collapses different documents with almost identical vectors. Metrics show that the patch level detects semantic changes, and confirm that aggregation obscures these details. We identify global texture dominance as the root cause. Our findings are consistent across model scales, retrieval-optimized embeddings, and multiple mitigation strategies, highlighting significant risks for single-vector visual document retrieval in financial applications.

Visual Document Retrieval (VDR) models mostly rely on late interaction architectures, in which documents are represented by a set of local patch embeddings and then matched against query tokens. While efficient, this architecture prioritizes local similarity over global layout structure of documents to estimate relevancy between documents and query. In practice, this leads to errors as relevance originates from layout structure of documents with heterogeneous layouts combining figures, tables, and text. We make document layout learnable without changing inference. We propose a multimodal encoder that augments local patch representations with a global layout embedding, trained via textual descriptions encoding document layout information. Across four ViDoRe-v2 datasets, our model improves over the strongest architecturally comparable ColPali/ColQwen baseline by +2.4 nDCG@5 and +2.3 MAP@5, with statistically significant per-dataset gains over ColQwen.

Multimodal Large Language Models (MLLMs) have achieved remarkable performance in Visually Rich Document Understanding (VRDU) tasks, but their capabilities are mainly evaluated on pristine, well-structured document images. We consider content restoration from shredded fragments, a challenging VRDU setting that requires integrating visual pattern recognition with semantic reasoning under significant content discontinuities. To facilitate systematic evaluation of complex VRDU tasks, we introduce ShredBench, a benchmark supported by an automated generation pipeline that renders fragmented documents directly from Markdown. The proposed pipeline ensures evaluation validity by allowing the flexible integration of latest or unseen textual sources to prevent training data contamination. ShredBench assesses four scenarios (English, Chinese, Code, Table) with three fragmentation granularities (8, 12, 16 pieces). Empirical evaluations on state-of-the-art MLLMs reveal a significant performance gap: The method is effective on intact documents; however, once the document is shredded, restoration becomes a significant challenge, with NED dropping sharply as fragmentation increases. Our findings highlight that current MLLMs lack the fine-grained cross-modal reasoning required to bridge visual discontinuities, identifying a critical gap in robust VRDU research.

Recent advances in vision-language models have demonstrated remarkable performance across diverse multi-modal tasks, including document question answering that leverages structured visual cues from text, tables, and figures. However, unlike natural images, document images contain large backgrounds and only sparse supporting evidence, leading to the inefficient consumption of substantial computational resources, especially for long documents. We observe that existing token-reduction methods for natural images and videos fall short in utilizing the structural sparsity unique to documents. To address this, we propose DocPrune, a training-free and progressive document token pruning framework designed for efficient long-document understanding. The proposed method preserves only the essential tokens for the task while removing unnecessary ones, such as background or question-irrelevant tokens. Moreover, it automatically selects the appropriate layers to initiate token pruning based on the model's level of comprehension. Our experiments on the M3DocRAG show that DocPrune improves throughput by 3.0x and 3.3x in the encoder and decoder, respectively, while boosting the F1 score by +1.0, achieving both higher accuracy and efficiency without any additional training.

This paper presents a computational case study that evaluates the capabilities of specialized machine learning models and emerging multimodal large language models for Visual Political Communication (VPC) analysis. Focusing on concentrated visibility in Instagram stories and posts during the 2021 German federal election campaign, we compare the performance of traditional computer vision models (FaceNet512, RetinaFace, Google Cloud Vision) with a multimodal large language model (GPT-4o) in identifying front-runner politicians and counting individuals in images. GPT-4o outperformed the other models, achieving a macro F1-score of 0.89 for face recognition and 0.86 for person counting in stories. These findings demonstrate the potential of advanced AI systems to scale and refine visual content analysis in political communication while highlighting methodological considerations for future research.

Visual long-document understanding is critical for enterprise, legal, and scientific applications, yet the best performing open recipes have not explored reasoning, a capability which has driven leaps in math and code performance. We introduce a synthetic data pipeline for reasoning in long-document understanding that generates thinking traces by scoring each page for question relevance, extracting textual evidence and ordering it from most to least relevant. We apply SFT to the resulting traces within \texttt{<think>} tags, gated by a \texttt{<cot>} control token, and the resulting reasoning capability is internalized via low-strength model merging. We study Qwen3 VL 32B and Mistral Small 3.1 24B. With Qwen3 VL, we achieve 58.3 on MMLongBenchDoc, surpassing the 7$\times$ larger Qwen3 VL 235B A22B (57.0). With Mistral, we show that synthetic reasoning outperforms distillation from the Thinking version's traces by 3.8 points on MMLBD-C, and internalized reasoning exhibits 12.4$\times$ fewer mean output tokens compared to explicit reasoning. We release our pipeline for reproducibility and further exploration.

With the rapid proliferation of multimodal information, Visual Document Retrieval (VDR) has emerged as a critical frontier in bridging the gap between unstructured visually rich data and precise information acquisition. Unlike traditional natural image retrieval, visual documents exhibit unique characteristics defined by dense textual content, intricate layouts, and fine-grained semantic dependencies. This paper surveys the VDR landscape as a retrieval problem in its own right, rather than as the front end of a generation pipeline, and does so specifically through the lens of the Multimodal Large Language Model (MLLM) era. We begin by examining the benchmark landscape, including the recent turn toward reasoning-intensive evaluation, and then dive into the methodological evolution along two orthogonal axes: what a retriever is, spanning multimodal embedding models and reranker models, and how it is deployed, from single-stage retrieval through Retrieval-Augmented Generation (RAG) to Agentic systems. Cutting across both, we analyse the representation--efficiency trade-off that late interaction imposes. We ground these categories in leaderboard evidence on where the empirical frontier actually lies, and close by identifying persistent challenges and outlining promising future directions for multimodal document intelligence.

We present the first comprehensive, large-scale study of training long-context vision language models up to 344K context, targeting long-document visual question answering with measured transfer to long-context text. While several such strong are open-weight, namely Qwen3 VL and GLM 4.5/6V, their training recipes and data pipelines are not reproducible. We systematically study continued pretraining, supervised finetuning, and preference optimization for 24B and 32B parameter models, backed by extensive LC evaluations and ablations to bridge this gap, and achieve state-of-the-art performance on MMLongBenchDoc for both parameter scales. In addition to this, our key findings include: (i) training on context lengths that match evaluation context lengths outperforms training on longer contexts, (ii) training and evaluating with page indices provides a simple, high-impact boost to long-document performance, (iii) our synthetic data pipelines enable self-improvement via continued pretraining and supervised finetuning, and (iv) we extend the known text-to-visual long context transfer to the reverse, showing that visual long context training transfers to long-context text performance. We also release MMLBD-C, a manually corrected version of MMLongBenchDoc to reduce erroneous and low quality examples in the benchmark.