Visual Document Understanding and Reasoning: A Multi-Agent Collaboration Framework with Agent-Wise Adaptive Test-Time Scaling

TL;DR

This paper introduces MACT, a multi-agent framework for visual document understanding that dynamically allocates computational resources through adaptive test-time scaling, improving reasoning accuracy and efficiency.

Contribution

It pioneers a procedural scaling paradigm with a multi-agent architecture and adaptive resource allocation for visual document reasoning tasks.

Findings

Achieves top-three performance on multiple benchmarks.

Improves accuracy by 9.9-11.5% over base models.

Uses fewer parameters while maintaining reasoning capabilities.

Abstract

The dominant paradigm of monolithic scaling in Vision-Language Models (VLMs) is failing for understanding and reasoning in documents, yielding diminishing returns as it struggles with the inherent need of this domain for document-based procedural reasoning, cognitive complexity, and factual accuracy. To this end, we introduce MACT, a Multi-Agent Collaboration framework with agent-wise adaptive Test-time scaling that pioneers a paradigm shift to procedural scaling, adapting dynamically to the functional entities of visual documents understanding and reasoning. MACT decomposes the visual document processing flow into four specialized agents, i.e., planning, execution, judgment, and answer, to resolve cognitive overload and introduce a critical self-correction loop for factual grounding. This collaborative architecture is amplified by an agent-wise adaptive test-time scaling strategy that intelligently allocates computational resources based on the complexity and redundancy of each functionality. Evaluated on multiple visual document understanding benchmarks, MACT achieves superior performance with a smaller parameter scale, adapting effectively to various document scenarios without compromising its general or mathematical reasoning capabilities. The three variants of MACT consistently attain top-three average performance rankings, with average performance enhancements of 9.9-11.5% over the base models. The source code will be released publicly.