通过并行标记预测实现高效的文档解析
Efficient Document Parsing via Parallel Token Prediction
AI总结：
本文提出并行标记预测方法，通过在输入序列中插入可学习标记并设计训练目标，使VLM在文档解析中实现并行解码，提升解析速度并减少模型幻觉。
摘要：
文档解析作为一项基础且关键的视觉任务，正被视觉-语言模型（VLMs）所革新。然而，VLMs固有的自回归（AR）解码特性成为显著瓶颈，严重限制了解析速度。本文提出并行标记预测（PTP），一种可插拔、模型无关且简单有效的技术，使VLM能够并行生成多个未来标记，提高采样效率。具体而言，我们将在输入序列中插入一些可学习的标记，并设计相应的训练目标，使模型具备文档解析的并行解码能力。此外，为支持有效训练，我们开发了一套全面的数据生成管道，高效地生成大规模、高质量的文档解析训练数据。在OmniDocBench和olmOCR-bench上的大量实验表明，我们的方法不仅显著提高了解码速度（1.6x-2.2x），还减少了模型幻觉，并表现出强大的泛化能力。
英文摘要：
Document parsing, as a fundamental yet crucial vision task, is being revolutionized by vision-language models (VLMs). However, the autoregressive (AR) decoding inherent to VLMs creates a significant bottleneck, severely limiting parsing speed. In this paper, we propose Parallel-Token Prediction (PTP), a plugable, model-agnostic and simple-yet-effective method that enables VLMs to generate multiple future tokens in parallel with improved sample efficiency. Specifically, we insert some learnable tokens into the input sequence and design corresponding training objectives to equip the model with parallel decoding capabilities for document parsing. Furthermore, to support effective training, we develop a comprehensive data generation pipeline that efficiently produces large-scale, high-quality document parsing training data for VLMs. Extensive experiments on OmniDocBench and olmOCR-bench demonstrate that our method not only significantly improves decoding speed (1.6x-2.2x) but also reduces model hallucinations and exhibits strong generalization abilities.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
