HybriDLA: Hybrid Generation for Document Layout Analysis

2025-11-24

AI Summary

Modern document layouts exhibit highly variable element counts and increasingly complex structures, causing performance degradation in conventional layout analysis methods. To address this, we propose a generative document layout analysis framework that synergistically integrates diffusion modeling with autoregressive decoding. To our knowledge, this is the first approach to jointly model iterative bounding-box refinement (via diffusion) and semantic-context-aware sequence generation (via autoregressive decoding) within a single, unified architecture. We further introduce a multi-scale feature fusion encoder to jointly capture fine-grained localization cues and high-level semantic structural information. Evaluated on DocLayNet and M⁶Doc benchmarks, our method achieves a new state-of-the-art mAP of 83.5%, substantially outperforming existing approaches. These results demonstrate the effectiveness and generalizability of the generative paradigm for complex document understanding.

Technology Category

• Natural Language Processing: Generation
• Machine Learning: Deep Generative Models & Autoencoders
• Computer Vision: Diffusion Models for Vision

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG

Abstract

Conventional document layout analysis (DLA) traditionally depends on empirical priors or a fixed set of learnable queries executed in a single forward pass. While sufficient for early-generation documents with a small, predetermined number of regions, this paradigm struggles with contemporary documents, which exhibit diverse element counts and increasingly complex layouts. To address challenges posed by modern documents, we present HybriDLA, a novel generative framework that unifies diffusion and autoregressive decoding within a single layer. The diffusion component iteratively refines bounding-box hypotheses, whereas the autoregressive component injects semantic and contextual awareness, enabling precise region prediction even in highly varied layouts. To further enhance detection quality, we design a multi-scale feature-fusion encoder that captures both fine-grained and high-level visual cues. This architecture elevates performance to 83.5% mean Average Precision (mAP). Extensive experiments on the DocLayNet and M$^6$Doc benchmarks demonstrate that HybriDLA sets a state-of-the-art performance, outperforming previous approaches. All data and models will be made publicly available at https://yufanchen96.github.io/projects/HybriDLA.

Problem

• Addresses limitations of conventional document layout analysis with modern complex layouts
• Unifies diffusion and autoregressive decoding for precise region prediction
• Enhances detection quality using multi-scale feature-fusion encoder

Innovation

• Hybrid diffusion-autoregressive decoding for layout analysis
• Multi-scale feature-fusion encoder captures visual cues
• Unified framework refines bounding-box hypotheses iteratively