Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

DocCogito: Aligning Layout Cognition and Step-Level Grounded Reasoning for Document Understanding

Yuchuan Wu, Minghan Zhuo, Teng Fu, Mengyang Zhao, Bin Li, Xiangyang Xue

TL;DR

DocCogito is proposed, a unified framework that integrates global layout perception with structured, region-grounded reasoning and introduces a lightweight layout tower that distills page structure into learnable global layout prior tokens, and a deterministic Visual-Semantic Chain to supervise fine-grained intermediate reasoning aligned with evidence regions.

Abstract

Document understanding with multimodal large language models (MLLMs) requires not only accurate answers but also explicit, evidence-grounded reasoning, especially in high-stakes scenarios. However, current document MLLMs still fall short of forming a complete, human-like reasoning process, because even when they improve both layout encoding and CoT-style prompting, the interaction between the two is typically learned implicitly and remains loosely coupled rather than being enforced as a systematic mechanism. So we propose DocCogito, a unified framework that integrates global layout perception with structured, region-grounded reasoning. DocCogito introduces a lightweight layout tower that distills page structure into learnable global layout prior tokens, and a deterministic Visual-Semantic Chain (VSC)-a concise structured representation less ambiguous than free-form natural-language CoT-to supervise fine-grained intermediate reasoning aligned with evidence regions. Training follows a progressive recipe, including layout perception pretraining, VSC-guided cold start, rejection sampling, and GRPO. To further strengthen the internal coupling between layout priors and VSC execution, we augment standard rewards with a fine-grained region-confidence signal that encourages reasoning traces to stay aligned with corresponding evidence regions. Extensive experiments on six benchmarks (DocVQA, WTQ, ChartQA, TextVQA, OCRBench, and InfoVQA) demonstrate strong generalization, achieving state-of-the-art results on four benchmarks.

DocCogito: Aligning Layout Cognition and Step-Level Grounded Reasoning for Document Understanding

TL;DR

DocCogito is proposed, a unified framework that integrates global layout perception with structured, region-grounded reasoning and introduces a lightweight layout tower that distills page structure into learnable global layout prior tokens, and a deterministic Visual-Semantic Chain to supervise fine-grained intermediate reasoning aligned with evidence regions.

Abstract

Document understanding with multimodal large language models (MLLMs) requires not only accurate answers but also explicit, evidence-grounded reasoning, especially in high-stakes scenarios. However, current document MLLMs still fall short of forming a complete, human-like reasoning process, because even when they improve both layout encoding and CoT-style prompting, the interaction between the two is typically learned implicitly and remains loosely coupled rather than being enforced as a systematic mechanism. So we propose DocCogito, a unified framework that integrates global layout perception with structured, region-grounded reasoning. DocCogito introduces a lightweight layout tower that distills page structure into learnable global layout prior tokens, and a deterministic Visual-Semantic Chain (VSC)-a concise structured representation less ambiguous than free-form natural-language CoT-to supervise fine-grained intermediate reasoning aligned with evidence regions. Training follows a progressive recipe, including layout perception pretraining, VSC-guided cold start, rejection sampling, and GRPO. To further strengthen the internal coupling between layout priors and VSC execution, we augment standard rewards with a fine-grained region-confidence signal that encourages reasoning traces to stay aligned with corresponding evidence regions. Extensive experiments on six benchmarks (DocVQA, WTQ, ChartQA, TextVQA, OCRBench, and InfoVQA) demonstrate strong generalization, achieving state-of-the-art results on four benchmarks.
Paper Structure (19 sections, 13 equations, 5 figures, 4 tables)

This paper contains 19 sections, 13 equations, 5 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Document Understanding
  4. Reinforcement Learning
  5. Methodology
  6. Dataset Construction
  7. Visual-Semantic Chain (VSC)
  8. Model Architecture
  9. Training Recipe
  10. Layout Perception Pretraining
  11. Multi-stage Post-training
  12. Reward Formulation
  13. Experiments
  14. Datasets and Metrics
  15. Implementation Details
  16. ...and 4 more sections

Figures (5)

  • Figure 1: Overview of reasoning pipelines in document understanding models. (a) Traditional text-only CoT. (b) DocLayLLM with OCR-based text/boxes and multiple CoT templates. (c) LayoutLLM with OCR-based text/boxes and staged reasoning (question analysis $\rightarrow$ area concentration $\rightarrow$ answer formation). (d) Ours, an OCR-free approach that integrates global layout perception with structured, region-grounded reasoning. VSC means Visual-Semantic Chain.
  • Figure 2: Overview of our methods. (a) The model architecture, where a lightweight layout tower extracts global layout cues and injects a [LAYOUT] token into the LLM. (b) The two-stage training recipe, consisting of layout perception pretraining and multi-stage GRPO.
  • Figure 3: OCRBench subtask breakdown for the top three overall models (Marten, Mini-Monkey, and DocCogito-8B) under the official evaluation protocol.
  • Figure 4: A VSC-style CoT example showing question analysis, region grounding, and operator-level reasoning steps that lead to the final answer.
  • Figure 5: Qualitative examples of our VSC-style CoT across diverse document types.