document digitization teams

Abstract: Discrete visual tokens represent vision in a discrete form analogous to language, offering a native interface for multimodal modeling under a shared next-token prediction objective. However, quantization error is widely assumed to cause unavoidable semantic loss, fundamentally limiting the semantic fidelity of discrete visual tokens and positioning continuous visual representations as an intrinsic upper bound. Using multimodal understanding as a proxy for semantic fidelity, we revisit this assumption and find quantization error unreliable, as discrete representations with substantially larger errors can achieve better understanding. Instead, the effective bit budget of discrete indices emerges as a key determinant of downstream performance. However, existing quantizers eventually saturate as additional codebooks contribute progressively less semantic information. As one approach to alleviating this saturation, we introduce Factorized Vector Quantization, which quantizes complementary, fixed-dimensional views of the visual feature with symmetric parallel codebooks. It continues to improve at larger bit budgets, ultimately matching continuous visual representations in multimodal understanding. Trained on industrial-scale multimodal data, our model achieves state-of-the-art performance among discrete-representation models, outperforming UniAR and Emu3-Chat by 22.8% and 59.2%, respectively, on a nine-metric average. To our knowledge, it is also the first discrete-representation model competitive with strong continuous-representation models such as Qwen3-VL, including on fine-grained OCR tasks. We further identify the key principles and a practical recipe enabling this competitiveness. These results and insights support the potential of discrete visual representations to underpin large-scale native multimodal foundation models. Code and models are available at https://github.com/hustvl/FVQ

Abstract: Cross-modal translation requires models to capture shared semantics across modalities while preserving the fine-grained structure required by the target modality. Yet in real-world scientific domains, paired data are often imperfect, while obtaining clean pairs at scale is costly. Supervised fine-tuning (SFT) directly optimizes cross-modal translation through categorical token-level supervision, but this objective is sensitive to errors in the target sequences. JEPA offers a complementary representation-space objective, but most existing formulations focus on prediction within a modality rather than translation across modalities. We introduce Anchored Cross-Modal JEPA (AC-JEPA), an SFT-compatible plug-in that predicts masked intermediate states in both cross-modal directions using paired data. The joint-embedding objective targets cross-modal structure predictable from context, while the causal SFT path retains direct token-level supervision for the target sequence. During self-training, AC-JEPA learns from continuous latent targets derived from imperfect pseudo-labeled pairs, offering intrinsic tolerance to local data errors without promoting pseudo-tokens to hard categorical supervision. Using optical chemical structure recognition as a testbed, AC-JEPA trained on clean pairs achieves 76.07% accuracy on the Simple set of the WildMol benchmark, compared with 67.40% for SFT alone. Using the same pseudo-labeled data for both approaches, AC-JEPA-based self-training further raises accuracy to 80.23%, whereas accuracy under direct SFT drops from 67.40% to 63.47%. These results support SFT-anchored joint-embedding prediction as an effective strategy for cross-modal translation under both curated and imperfect supervision.

Abstract: Optical character recognition (OCR) systems depend strongly on the scale and diversity of their training data. One approach to document OCR first analyzes the page layout and extracts text-line images, which are then transcribed by a recognition model. Multilingual line-level recognition involving Persian, Arabic, and English presents additional challenges because different writing directions, fonts, and styles may coexist within a single line. This paper introduces MuSTEL, a large-scale synthetic multilingual OCR dataset containing more than 14.5 million annotated text-line images, organized into a training partition and predefined small and large test sets. Samples are generated from natural-language corpora, including multilingual content already present in the source texts. The generation framework incorporates capability-aware font selection, within-line font and style changes, mixed-direction text rendering, highlighting, and underlining. A validated repository of 2,009 font files from 711 families supports broad typographic coverage, while varied backgrounds, blur, geometric distortion, skew, stroke width, and image save quality provide diverse document-like appearances. Statistical analysis characterizes the dataset's language composition, transcription lengths, lexical and character-sequence diversity, image dimensions, typography, and rendering conditions. MuSTEL provides a resource for training and evaluating multilingual line-level OCR systems and investigating recognition robustness across languages, writing directions, and visual conditions.

Abstract: Multimodal large language models (LLMs) are increasingly used for structured extraction from financial documents, but their behavior under visual ambiguity is poorly understood. We study CARLAR (Courtesy Amount Recognition and Legal Amount Recognition) influence: a model returns the same amount for both cheque fields even though their annotations differ, including cases with a blank annotation encoded as zero. Across ten visual prompting conditions, we evaluate GPT-4.1 and Claude 4.5 Sonnet on 243 primary cheques (124 with two observed unequal amounts) and 598 strict mismatches from two additional datasets. Under courtesy redaction, GPT-4.1 CARLAR reaches 73.7 percent (95 percent bootstrap interval: 68.3 to 79.0; p below 0.001 versus baseline). Model-reported confidence can remain high when predictions agree; expected calibration error reaches 0.93 among CARLAR cases. We also test a compressed-image few-shot intervention that presents examples as one visual table without model fine-tuning. On the 243-case cohort, Claude's Control condition increases paired joint extraction accuracy by 11.5 percentage points (95 percent interval: 6.2 to 16.9; adjusted p = 0.00054), but its strict-124 gain is smaller and not significant after correction. The intervention's effect on predicted-field agreement is model-dependent: it decreases for Claude and increases for GPT-4.1. These results motivate validation methods that account for visual context and dependencies between extracted fields.

Abstract: Distinguishing visually similar characters is a persistent challenge for optical character recognition (OCR) systems, causing errors in critical applications of text digitization. While preprocessing techniques improve image quality and word-level post-processing corrects spelling errors, neither addresses character-level visual ambiguity, an intrinsic property of certain character pairs that limits even state-of-the-art OCR models. We present the Character Verification Model (CVM), a character-level post-processing framework that verifies and corrects ambiguous characters in OCR outputs using specialized binary classifiers trained on visually confusable pairs such as 'O'/'0', 'Z'/'2', and 'S'/'5'. CVM operates independently of dictionary validation, making it suitable for non-lexical data including serial numbers, alphanumeric codes, and license plates. Experiments on three OCR engines (Tesseract, EasyOCR, TrOCR) and ten fonts with low to high complexity demonstrate accuracy improvements of 3.95-14.81 percentage points, with an 87.3% correction rate for visually similar character errors and a false-correction rate of only 2.3%. The modular architecture integrates with existing OCR pipelines without modification to underlying recognition models.

Abstract: Diffusion models have achieved remarkable success in image super-resolution (SR) by leveraging strong generative priors to enhance perceptual quality. However, for textual content, these priors often produce text hallucinations: non-existent or distorted text, which alter scene semantics and compromise model reliability. While recent methods focus on enhancing text generation, they do so indiscriminately, leaving hallucinations fundamentally unresolved. We argue the core problem is a lack of selective control. Therefore, we reformulate text SR as a selective generation problem: text is resolved only when reliable generation is expected, while other scene content is fully super-resolved. We introduce TGSR (Text-Grounded Super-Resolution), a test-time framework that progressively identifies and suppresses likely hallucinated text regions during generation. To enable this, we train dedicated detectors on a large-scale dataset curated from diverse diffusion-based SR outputs. TGSR consistently improves text fidelity across multiple backbones, increasing recognition precision by up to 17.68 points without degrading perceptual quality, underscoring the value of selective control in generative SR.

Abstract: Intelligent grading and automated scoring technologies constitute critical infrastructure for smart education. However, existing document parsing and handwriting recognition benchmarks are predominantly designed for well-structured printed documents or isolated mathematical expressions, lacking datasets that capture the complex characteristics inherent to student answer sheets, including multi-line derivation processes, heterogeneous mixtures of text and mathematical formulae, and noise artifacts such as strikethroughs. To address this gap, we introduce HANS, the first dataset explicitly constructed for real-world educational scenarios, encompassing mathematical expressions, natural language text, hand-drawn tables, and diverse noise patterns including corrections and deletions, accompanied by fine-grained annotations that establish a reliable foundation for robust recognition research. Building upon HANS, we propose NA-GOT, an end-to-end framework that achieves two-stage noise suppression through a lightweight noise suppression module operating at the feature level, complemented by a noiseaware attention mechanism incorporated into the decoding stage. Experimental results demonstrate that HANS poses substantial challenges to existing methods, while NA-GOT achieves significant improvements in both accuracy and stability for answer process recognition. The dataset will be made publicly available upon publication.

Abstract: Zero-shot Chinese character recognition (ZS-CCR) aims to recognize characters whose categories are never observed during training, and typically relies on the compositional structure shared between seen and unseen characters. Recent CLIP-style methods represent this structure with the Ideographic Description Sequence (IDS) and align it with glyph images in a shared embedding space. However, they rely on a single global image--IDS similarity that discards the spatial layout of radicals and, being learned only implicitly from seen classes, generalizes poorly to unseen ones; moreover, global matching often retrieves the correct character within the top candidates yet fails to rank it first when characters differ only in subtle local radicals. To address these issues, we propose a global-to-local two-stage framework. In the first stage, STG-CLIP augments the IDS with explicit tree-position and radical-level geometric priors, yielding a spatial-aware prototype that provides a consistent spatial description across seen and unseen categories for high-recall global retrieval. In the second stage, the Radical Verification Module (RVM) uses the radical instances of each retrieved candidate as queries to verify whether the corresponding radicals can be matched to spatially compatible regions in the input glyph. A margin-based gating rule activates the RVM only when the leading global candidates receive similar similarity scores. Experiments on the ICDAR2013 benchmark demonstrate that our method achieves state-of-the-art performance under the character-level zero-shot setting, obtaining 83.06% top-1 accuracy with 2,755 seen classes. Ablation studies further show that the explicit geometric priors and radical-level verification provide complementary improvements.

Abstract: Large Vision-Language Models achieve strong VQA performance, but processing high-resolution, information-rich images requires substantial computation, motivating visual token reduction. However, existing methods often prune individual tokens or rely on fixed-size cropping, limiting their ability to preserve spatially structured information such as horizontally or vertically elongated text. To address this limitation, we propose ReFIT, an instruction-guided visual token reduction framework for efficient LVLM inference. ReFIT consists of Relevance-Guided Window Reshaping (RWR) and Instruction-Guided Token Refinement (ITR), where RWR captures instruction-relevant regions by adapting to their spatial characteristics, while ITR further removes unnecessary visual tokens. Experiments on four VQA benchmarks demonstrate that ReFIT improves answer accuracy while reducing computational cost, and qualitative results demonstrate its effectiveness in localizing relevant regions and removing unnecessary visual information.

Abstract: Direct visual readability and task-information recoverability are different quantities. Failure to decode a target from a fixed observation need not eliminate access to that target through another recovery route. We develop an evaluation perspective that makes the observation, query, target, and available knowledge explicit and measures the overlap between routes' success sets. For information available on the original visible surface under suitable imaging conditions, direct optical recovery reads the target from the image, optionally after restoration; entity-linked recovery uses residual visual evidence to identify the depicted entity and accesses its target through an entity--attribute relation in a specified knowledge resource. Such access can draw on stored knowledge or an external source. A controlled book-cover study instantiates external access with a fixed title--author catalog, comparing optical author recovery with visual title resolution and deterministic lookup under resolution degradation. Entity-linked successes persist across the tested vision--language models, revealing information access beyond the tested direct visual frontier despite substantial differences in absolute performance. A substantial optical-only region remains. These complementary outcomes show why visual degradation should be evaluated through the task information accessible along specified routes and knowledge resources, alongside direct readability.

Abstract: In self-supervised pretraining for Handwritten Text Recognition (HTR), pixel reconstruction methods outperform contrastive methods, unlike in natural-image classification. We argue that this difference follows from where discriminative signal lies in pixel space: for HTR, it is concentrated in high-variance directions and largely absent from low-variance ones. This predicts that objectives preserving high-variance pixel content will transfer best. We test six SSL methods from three families (pixel-grounded MIM, JEPA, and contrastive) under matched encoder, data, and evaluation protocols on six handwriting benchmarks across five languages. With full labels, pixel-groundrounded SSL achieves the lowest CER on every benchmark and both frozen probes, exposes per-position character information that other families recover only through the readout, and is the only family to benefit from pretraining on real handwriting. Pixel-grounded representations are also more label efficient. Across datasets, encoder alignment with the high-variance pixel subspace predicts CER within every method. With a pretrained LLM decoder, a frozen pixel-grounded encoder is competitive with fully fine-tuned supervised baselines; full fine-tuning achieves the lowest mean CER and ranks first or second on every benchmark. These results show that the value of pixel reconstruction depends on where discriminative signal lies in the input.

Abstract: Vision-language models (VLMs), despite their success in optical character recognition (OCR) tasks, are vulnerable to typographic attacks and have a fragile structure for images with multiple text layers. In this study, the DecoyBench dataset was created using the Decoy Font method. The dataset consists of 300 images, each containing text with sharp contour lines superimposed on another text with soft shading. Six recent closed-source models from three different model families were evaluated using this dataset under two different prompting conditions (naive and guided) and at two different resolutions ($512\times512$ and $64\times64$). A validation study showed that human participants could read both text layers with high accuracy. In contrast, the models, with most variants and both prompting methods, read the contour text with near-human accuracy at high resolution, but almost never fully extracted the shading text. At low resolution, the contour text could not be read by either the models or humans, while the shading text could be extracted with high accuracy. The findings indicate that the evaluated VLMs exhibit a consistent behavioral limitation when processing typographic structures containing multiple spatial frequency layers.

Abstract: Multilingual scene text recognition (STR) remains challenging due to the scarcity of training data for most languages and the difficulty of serving diverse scripts within a single model. Existing solutions either deploy one recognizer per language, inflating cost and introducing error accumulation, or rely on massive vision-language models (VLMs) that are expensive and still inaccurate on many scripts. In this work, we pursue an all-in-one multilingual recognizer that is simpler than per-language experts, lighter than VLMs, and more accurate than both. First, we construct TextMuSS-10M, a large-scale synthetic scene text dataset spanning 10 scripts and 229 languages. It provides balanced and sufficient supervision where real data is unavailable. Second, we propose ScriptMoE, a script-aware Mixture-of-Experts (MoE) architecture. It shares a single visual encoder and replaces the dense decoder with a sparse MoE block, which consists of an image-level router dispatches each image to the top-2 script-aligned experts and a shared expert absorbs cross-script knowledge. Extensive experiments on our assembled TextMuSS-Bench (10 scripts, 10,899 images) show that ScriptMoE achieves the highest accuracy of 82.06%, outperforming the strongest STR baseline by 1.31%. On the CC-OCR end-to-end multilingual task, replacing only the recognizer in PP-OCRv5 with ScriptMoE lifts F1 score from 65.71% to 80.89%, slightly surpassing the best VLM (80.73%) at a fraction of the parameter count.

Abstract: Table extraction suffers from frequent structural errors and semantic hallucinations. We propose PrismAlign, a multi-VLM framework aligning diverse visual perspectives to resolve ambiguity. It integrates priors of table logic to assess output plausibility, decoupling structural alignment from cell content alignment. A Bayesian decision strategy maximizes alignment accuracy by exploiting the correlation between extraction errors and computable rule violations. Evaluated on open-source and custom VLMs, PrismAlign reduces hallucinations and achieves state-of-the-art performance on OmniDocBench 1.5, as well as on the table category of CC-OCR and PureDocBench.

Abstract: Recovering numerical series from line plots requires accurate axis calibration and reliable curve extraction. We present LinePilot Digitizer (LinePilot), which combines continuous color-based curve recovery with three calibration modes: LinePilot (standard), LinePilot (enhanced), and LinePilot (OCR). We also introduce DigitizerBench, the first dedicated benchmark for systematically evaluating digitizer performance, using an orthogonal design spanning signal, rendering, and plot-structure factors with complementary automatic and human-guided evaluations. We evaluate performance using failure-penalized capped normalized root-mean-square error (FPC-NRMSE), which assigns unit loss to missing, unusable, or catastrophically inaccurate outputs. On DigitizerBench-Full, LinePilot (OCR) achieves the lowest mean FPC-NRMSE (0.672) and highest trusted usability (38.2%) among the tested automatic pipelines. On DigitizerBench-Lite, LinePilot (enhanced) achieves the lowest mean FPC-NRMSE (0.081), 100% output success, and highest trusted usability (93.3%). The orthogonal benchmark design further enables factor analysis to identify the factors that most significantly affect digitizer performance. Together, the three calibration modes provide a practical trade-off between automation, user control, and accuracy within a shared curve-recovery workflow.

Abstract: Text-centric Visual Question Answering (VQA) requires reading and reasoning over text embedded in images, a task made substantially harder when images suffer from real-world degradation such as motion blur, low resolution, or compression artifacts. While modular OCR-based pipelines and end-to-end vision-language models are both widely used for this task, their comparative robustness under degraded conditions remains underexplored. We present an empirical study comparing two modular pipelines with SA-DBNet, a custom detector architecture combining ResNet-18 with self-attention spatial modeling and deformable convolutions against an end-to-end vision-language baseline, evaluated on 4013 degraded images with 7000 question-answer pairs. Fine-tuned modular pipelines achieve up to 57.50% exact-match accuracy versus 38.00% for the end-to-end baseline, with domain-specific fine-tuning yielding a gain of up to 29.50 percentage points. Critically, we find that conventional OCR error metrics like Character Error Rate and Word Error Rate are unreliable predictors of downstream VQA performance, as semantic reasoning can compensate for recognition failures when contextual cues are present. These findings highlight the importance of task-aware evaluation for text-centric VQA systems under realistic visual conditions. Codes are available here

Abstract: Handwritten mathematical expression recognition (HMER) is conventionally scored by exact-match rates and string-similarity metrics that are blind to where an error occurs: two predictions with identical token-error counts receive identical scores whether they misplace a subscript or swap the operands of a fraction. Render-based character detection matching (CDM) aligns glyphs robustly but remains position-blind---on controlled fraction-operand swaps it scores 0.8595 where position-aware scoring yields 0.6253. Tree-edit metrics exhibit a complementary blind spot: rewrites outside the parser's normalization coverage are penalized as structural errors (0.8552 where render-based metrics score 1.0). We propose PA-CDM, a position-aware metric that couples character detection matching with position-forest encoding and divergence-level weighting; StructPerturb v2.0, a frozen benchmark of 1,340 controlled perturbation pairs across 15 type--intensity cells; and a cross-metric consistency protocol combining a sensitivity matrix, a human study, and LLM-judge calibration. In a six-annotator study, PA-CDM attains the highest correlation with human judgments among seven automatic metrics (Spearman rho=0.9535, n=990). A frontier LLM judge correlates slightly higher (rho=0.9613) but is costly, nondeterministic, and API-dependent; PA-CDM approaches it at zero marginal cost with deterministic, diagnosable behavior.

Abstract: Handwritten text recognition resources are often small and distributed across collections that differ in language, script, document structure, and annotation format, making joint page-level training difficult. We propose ExpertHTR, a unified vision-language framework that addresses this problem through complementary supervision and conditional model capacity. Structural annotations from heterogeneous datasets are first organized through a common Page-Region-Line representation and used to construct four related training tasks for complete transcription, physical-line coverage, text localization, and localized recognition, without requiring additional manual labels. Building on a jointly trained dense model, ExpertHTR introduces a sparse Mixture-of-Experts architecture with an always-active shared branch and conditionally routed full-MLP experts. Sparsegen allows the number of active routed experts to vary with the hidden representation, while routing regularization reduces persistent concentration on a small subset of experts. Experiments on seven heterogeneous handwriting benchmarks show that complementary supervision consistently improves training with page transcription alone, while joint multi-source training provides further gains on most datasets. The proposed sparse expert model further improves the dense baseline on six of the seven sources. The final unified model also substantially outperforms the evaluated general-purpose OCR and vision-language systems on most benchmarks and achieves state-of-the-art performance on the IAM paragraph-level benchmark, while specialized HTR systems remain stronger on several challenging collections.