让文档解析模型并行生成多个词，速度提升1.6倍以上
Efficient Document Parsing via Parallel Token Prediction
- 在输入中加入可学习标记，让模型学会并行预测多个未来词
- 在两个基准上实现1.6倍到2.2倍的解码加速，减少幻觉
- 适合需要高速文档解析的应用场景，如智能表单处理
文档解析作为关键的视觉任务，正被视觉语言模型（VLM）革新。然而，VLM固有的自回归解码机制严重制约了解析速度。本文提出并行令牌预测（PTP），一种可插拔、与模型无关、简单而有效的方案，使VLM能够并行生成多个未来令牌，提升采样效率。具体而言，我们在输入序列中插入可学习令牌，并设计相应的训练目标，赋予模型并行解码能力。此外，为支持有效训练，我们构建了一个全面的数据生成管道，高效生成大规模、高质量的VLM文档解析训练数据。在OmniDocBench和olmOCR-bench上的大量实验表明，该方法不仅显著提升解码速度（1.6x-2.2x），还减少了模型幻觉，并展现出强泛化能力。
原文摘要 · Abstract (English)
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.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。