Efficient Document Parsing via Parallel Token Prediction

TL;DR

This paper introduces Parallel-Token Prediction, a method that enables vision-language models to decode multiple tokens simultaneously, significantly boosting document parsing speed and reducing errors.

Contribution

We propose a simple, model-agnostic parallel decoding technique with a training pipeline for large-scale data generation, enhancing speed and accuracy in document parsing.

Findings

Decoding speed improved by 1.6x to 2.2x.

Reduced model hallucinations.

Strong generalization on benchmarks.

Abstract

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.