[2602.23632] MMKG-RDS: Reasoning Data Synthesis via Deep Mining of Multimodal Knowledge Graphs

Computer Science > Artificial Intelligence arXiv:2602.23632 (cs) [Submitted on 27 Feb 2026] Title:MMKG-RDS: Reasoning Data Synthesis via Deep Mining of Multimodal Knowledge Graphs Authors:Lun Zhan, Feng Xiong, Huanyong Liu, Feng Zhang, Yuhui Yin View a PDF of the paper titled MMKG-RDS: Reasoning Data Synthesis via Deep Mining of Multimodal Knowledge Graphs, by Lun Zhan and 4 other authors View PDF HTML (experimental) Abstract:Synthesizing high-quality training data is crucial for enhancing domain models' reasoning abilities. Existing methods face limitations in long-tail knowledge coverage, effectiveness verification, and interpretability. Knowledge-graph-based approaches still fall short in functionality, granularity, customizability, and evaluation. To address these issues, we propose MMKG-RDS, a flexible framework for reasoning data synthesis that leverages multimodal knowledge graphs. It supports fine-grained knowledge extraction, customizable path sampling, and multidimensional data quality scoring. We validate MMKG-RDS with the MMKG-RDS-Bench dataset, covering five domains, 17 task types, and 14,950 samples. Experimental results show fine-tuning Qwen3 models (0.6B/8B/32B) on a small number of synthesized samples improves reasoning accuracy by 9.2%. The framework also generates distinct data, challenging existing models on tasks involving tables and formulas, useful for complex benchmark construction. The dataset and code are available at this https URL Subjects: Arti...