MinerU-Diffusion: Rethinking Document OCR as Inverse Rendering via Diffusion Decoding

TL;DR

MinerU-Diffusion introduces a diffusion-based approach to document OCR, replacing traditional autoregressive methods, leading to faster, more robust, and less language-dependent long-form document understanding.

Contribution

The paper presents a novel diffusion decoding framework for OCR that enables parallel processing and improves robustness over autoregressive models.

Findings

Achieves up to 3.2x faster decoding than autoregressive baselines.

Demonstrates improved robustness and reduced language dependence.

Performs well on the Semantic Shuffle benchmark.

Abstract

Optical character recognition (OCR) has evolved from line-level transcription to structured document parsing, requiring models to recover long-form sequences containing layout, tables, and formulas. Despite recent advances in vision-language models, most existing systems rely on autoregressive decoding, which introduces sequential latency and amplifies error propagation in long documents. In this work, we revisit document OCR from an inverse rendering perspective, arguing that left-to-right causal generation is an artifact of serialization rather than an intrinsic property of the task. Motivated by this insight, we propose MinerU-Diffusion, a unified diffusion-based framework that replaces autoregressive sequential decoding with parallel diffusion denoising under visual conditioning. MinerU-Diffusion employs a block-wise diffusion decoder and an uncertainty-driven curriculum learning strategy to enable stable training and efficient long-sequence inference. Extensive experiments demonstrate that MinerU-Diffusion consistently improves robustness while achieving up to 3.2x faster decoding compared to autoregressive baselines. Evaluations on the proposed Semantic Shuffle benchmark further confirm its reduced dependence on linguistic priors and stronger visual OCR capability.