Paper page - HunyuanOCR-1.5: Making Lightweight OCR VLMs Faster and Better

Source: https://huggingface.co/papers/2607.04884 Authors:

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

Abstract

HunyuanOCR-1.5 is a lightweight end-to-end vision-language model that enhances OCR capabilities through improved efficiency via DFlash and enhanced capability via Agentic Data Flow, achieving fast inference and broad task coverage.

We present HunyuanOCR-1.5, a lightweight end-to-endOCR-specializedvision-language model. HunyuanOCRunifies document parsing, text spotting, information extraction, text-image translation, and multi-image document understanding within a single end-to-end VLM. Building upon the lightweight architecture of HunyuanOCR-1.0, HunyuanOCR-1.5 does not redesign the backbone, but systematically improves both efficiency and capability. For efficiency, we adaptDFlashtoOCRdecoding, significantly reducing the latency of long structured outputs such as dense documents, tables, and formulas while preserving output distribution. Powered byDFlash, HunyuanOCR-1.5 achieves a 6.37x Transformerinference speedupand a 2.14x speedup under vLLM, delivering the fastest inference among lightweightOCRVLMs. For capability, we proposeAgentic Data Flow, an agent-driven data construction system that transforms model weaknesses into executable data requirements and autonomously performs material search, quality verification, and pipeline development. It substantially improves long-tail capabilities in ancient-scriptOCR, fine-grained chart and table parsing, multi-image text-centric QA, low-resource multilingual parsing, and document hallucination evaluation. HunyuanOCR-1.5 ranks among the top-tier end-to-endOCRsolutions on OmniDocBench v1.6 while achieving new performance milestones across these long-tail tasks. Combined with an upgradedpretrainingandpost-trainingrecipe, HunyuanOCR-1.5 further extends its capability in high-resolution, long-context, and multi-task scenarios. Experiments demonstrate faster inference, broaderOCRcapability coverage, and the deployment advantages of a lightweightend-to-end model. We will release the model weights and training code to support future research and real-worldOCRapplications.

View arXiv pageView PDFAdd to collection

Get this paper in your agent:

hf papers read 2607\.04884

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

Models citing this paper2

#### tencent/HunyuanOCR Image-Text-to-Text• 1B• Updated1 day ago • 505k • 764 #### Evan-613/HunyuanOCR Image-Text-to-Text• 1B• Updatedabout 5 hours ago

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2607.04884 in a dataset README.md to link it from this page.

Spaces citing this paper14

Browse 14 spaces citing this paper## Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.