Search papers, labs, and topics across Lattice.
This paper introduces HPD-Parsing, a novel approach to document parsing that combines global layout analysis with parallel content decoding, addressing the limitations of existing unified Vision-Language Models that rely on sequential token generation. By employing a Hierarchical Parallel Decoding paradigm, the method organizes document structure while concurrently decoding block-level content, significantly improving throughput. Experimental results demonstrate that HPD-Parsing achieves 4,752 tokens per second, outperforming the fastest existing models by 2.62 times and maintaining competitive accuracy, thus highlighting the efficiency of hierarchical parallel decoding in document parsing tasks.
HPD-Parsing achieves over 2.6 times the throughput of the fastest existing document parsers by leveraging hierarchical parallel decoding, revolutionizing how we approach document layout and content parsing.
Efficient teamwork typically combines global coordination with parallel execution, a principle not yet fully reflected in unified Vision-Language Model (VLM)-based document parsers. Existing unified parsers process an entire page jointly but generate its output through a single token-by-token autoregressive trajectory, creating a sequential bottleneck that grows with document length. Such full-page sequential generation overlooks a key property of document parsing: layout must be analyzed globally, whereas block content can be parsed in parallel. Based on this observation, we introduce HPD-Parsing, which replaces full-page autoregressive generation with a Hierarchical Parallel Decoding paradigm. A main layout branch organizes the overall document structure and dynamically assigns block-level content decoding to concurrent branches, while progressive multi-token prediction (P-MTP) further reduces the decoding steps within each branch. Experiments on public benchmarks show that HPD-Parsing achieves 4,752 tokens per second, delivering $2.62\times$ the throughput of the fastest existing document parsing model and $3.06\times$ that of the vanilla autoregressive baseline, while maintaining competitive parsing accuracy. These results establish hierarchical parallel decoding as an effective alternative to full-page autoregressive generation, opening a new direction for efficient unified document parsing.
