PaDoc: Layout-Grounded Parallel Decoding for Document Parsing

Hugging Face Daily Papers Papers
document-parsing layout-grounding parallel-decoding mllm efficiency end-to-end

Summary

PaDoc introduces a layout-grounded parallel decoding method for end-to-end document parsers, decoupling layout and content decoding to reduce decoding depth and improve throughput. It achieves state-of-the-art results on OmniDocBenchFull and significantly speeds up inference compared to sequential baselines.

End-to-end document parsers provide a unified interface, but serialize page layouts and regional contents into one autoregressive sequence. This formulation forces independent regions onto a decoding path whose length grows with the total content, whereas crop-based two-stage parsers expose region-level parallelism at the cost of repeated visual prefills and fragmented page context. To retain full-page context while removing dependencies, we propose PaDoc, a layout-grounded parser that treats the predicted layout as a branching structure over a shared page representation. Under a region-sufficiency assumption, we derive a prefix-conditioned factorization in which the layout stream and regional content branches advance concurrently, reducing the decoding depth to the longest layout-content path. We realize this factorization within a single MLLM: packed variable-length ancestor attention preserves the visibility under standard next-token training, while masked parallel decoding creates branches that the evaluated vLLM backend serves as concurrent requests with cache-resident shared-prefix reuse. On OmniDocBench Full, PaDoc attains an Overall layout F1 of 91.1 and, among end-to-end parsers, a top-tier Overall score of 94.24 together with the best Text Edit (0.038) and Formula CDM (95.59). On a 384-page subset and one A800 GPU, it is the fastest end-to-end parser at five concurrency levels, improving valid-page throughput by 67.4-118% and reducing P95 latency by 39.2-54.9% relative to a same-backbone Sequential SFT baseline. Code is available at https://github.com/Longin-Yu/Padoc
Original Article
View Cached Full Text

Cached at: 08/07/26, 05:55 AM

Paper page - PaDoc: Layout-Grounded Parallel Decoding for Document Parsing

Source: https://huggingface.co/papers/2608.06146 Published on Aug 6

·

Submitted byhttps://huggingface.co/Longin-Yu

Hao Yuon Aug 7

Authors:

,

,

,

,

,

,

,

,

,

Abstract

End-to-enddocumentparsersprovideaunifiedinterface,butserializepagelayoutsandregionalcontentsintooneautoregressivesequence.Thisformulationforcesindependentregionsontoadecodingpathwhoselengthgrowswiththetotalcontent,whereascrop-basedtwo-stageparsersexposeregion-levelparallelismatthecostofrepeatedvisualprefillsandfragmentedpagecontext.Toretainfull-pagecontextwhileremovingdependencies,weproposePaDoc,alayout-groundedparserthattreatsthepredictedlayoutasabranchingstructureoverasharedpagerepresentation.Underaregion-sufficiencyassumption,wederiveaprefix-conditionedfactorizationinwhichthelayoutstreamandregionalcontentbranchesadvanceconcurrently,reducingthedecodingdepthtothelongestlayout-contentpath.WerealizethisfactorizationwithinasingleMLLM:packedvariable-lengthancestorattentionpreservesthevisibilityunderstandardnext-tokentraining,whilemaskedparalleldecodingcreatesbranchesthattheevaluatedvLLMbackendservesasconcurrentrequestswithcache-residentshared-prefixreuse.OnOmniDocBenchFull,PaDocattainsanOveralllayoutF1of91.1and,amongend-to-endparsers,atop-tierOverallscoreof94.24togetherwiththebestTextEdit(0.038)andFormulaCDM(95.59).Ona384-pagesubsetandoneA800GPU,itisthefastestend-to-endparseratfiveconcurrencylevels,improvingvalid-pagethroughputby67.4-118%andreducingP95latencyby39.2-54.9%relativetoasame-backboneSequentialSFTbaseline.Codeisavailableathttps://github.com/Longin-Yu/Padoc

View arXiv pageView PDFGitHub2Add to collection

Get this paper in your agent:

hf papers read 2608\.06146

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2608.06146 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2608.06146 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2608.06146 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

HPD-Parsing: Hierarchical Parallel Document Parsing

Hugging Face Daily Papers

HPD-Parsing introduces a hierarchical parallel decoding paradigm for VLM-based document parsing, replacing full-page autoregressive generation to achieve 4,752 tokens per second throughput (2.62x faster than prior models) while maintaining competitive accuracy.

dots.ocr: Multilingual Document Layout Parsing in a Single Vision-Language Model

Papers with Code Trending

This paper presents dots.ocr, a unified Vision-Language Model that jointly learns layout detection, text recognition, and relational understanding for multilingual document layout parsing. It achieves state-of-the-art results on OmniDocBench and introduces the XDocParse benchmark spanning 126 languages.