MetaReason: Precise Interleaved Multimodal Reasoning via Editing Meta Information for Solving Geometry Problems

2026-08-14

AI Summary

This study addresses visual representation distortion and low rendering fidelity in geometric reasoning by proposing the MetaReason framework. This approach parses images into structured meta-information to synthesize high-fidelity visual states via meta-editing, integrating Supervised Fine-Tuning and Reinforcement Learning to enable precise interleaved multimodal reasoning. Furthermore, we construct the TutorGeo and ExamGeo datasets alongside a controllable visual augmentation method. Experimental results demonstrate that MetaReason significantly outperforms open-source models across multi-level difficulty benchmarks and achieves performance comparable to proprietary commercial systems. These findings confirm the framework’s effectiveness in enhancing the reliability of complex geometric reasoning tasks.

Abstract

Although visual reasoning is crucial for solving complex geometry tasks, existing vision-language models rely heavily on text-only reasoning. Some recent methods introduce intermediate visual states to facilitate reasoning, but they are often hindered by inaccurate geometric representations and low rendering fidelity, ultimately leading to unreliable outputs. To address these limitations, we propose MetaReason, a framework for multimodal reasoning in plane geometry that leverages structured meta-information to enable accurate auxiliary-line construction. The framework first parses geometric images into meta-information, performs controllable edits with predefined tools to synthesize high-fidelity visual states, and then conducts reasoning based on these augmented views. To support this framework, we construct TutorGeo, a comprehensive dataset containing 17k image-to-meta conversion samples, 60k text-only reasoning traces, and 60k interleaved multimodal reasoning traces. Using this dataset, we combine supervised fine-tuning and reinforcement learning to develop robust multimodal reasoning capabilities. We also introduce ExamGeo, a benchmark derived from real-world examination problems that enables systematic evaluation across varying difficulty levels. Experimental results demonstrate that MetaReason significantly outperforms existing open-source models and achieves competitive performance against proprietary models.

Problem

• Geometry Problem Solving
• Multimodal Reasoning
• Vision-Language Models
• Visual Reasoning
• Geometric Representation

Innovation

• MetaReason
• Interleaved Multimodal Reasoning
• Meta-information Editing
• TutorGeo
• Reinforcement Learning