![[2504.20426] RV-Syn: Rational and Verifiable Mathematical Reasoning Data Synthesis based on Structured Function Library](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2504.20426] RV-Syn: Rational and Verifiable Mathematical Reasoning Data Synthesis based on Structured Function Library
Summary
The paper introduces RV-Syn, a novel approach for synthesizing high-quality mathematical reasoning data using structured function libraries, enhancing the capabilities of Large Language Models (LLMs).
Why It Matters
As LLMs increasingly require robust reasoning capabilities, the RV-Syn method addresses the critical need for verifiable and logically sound mathematical problem generation. This advancement could significantly improve the training datasets for AI models, leading to better performance in mathematical reasoning tasks.
Key Takeaways
- RV-Syn constructs a structured function library for mathematical operations.
- It generates computational graphs that enhance problem generation and solution verifiability.
- The method outperforms existing data synthesis techniques, including human-generated problems.
- RV-Syn provides a scalable framework for creating high-quality reasoning datasets.
- The approach supports the advancement of LLMs in mastering complex mathematical reasoning.
Computer Science > Artificial Intelligence arXiv:2504.20426 (cs) [Submitted on 29 Apr 2025 (v1), last revised 14 Feb 2026 (this version, v3)] Title:RV-Syn: Rational and Verifiable Mathematical Reasoning Data Synthesis based on Structured Function Library Authors:Jiapeng Wang, Jinhao Jiang, Zhiqiang Zhang, Jun Zhou, Wayne Xin Zhao View a PDF of the paper titled RV-Syn: Rational and Verifiable Mathematical Reasoning Data Synthesis based on Structured Function Library, by Jiapeng Wang and Jinhao Jiang and Zhiqiang Zhang and Jun Zhou and Wayne Xin Zhao View PDF HTML (experimental) Abstract:The advancement of reasoning capabilities in Large Language Models (LLMs) requires substantial amounts of high-quality reasoning data, particularly in mathematics. Existing data synthesis methods, such as data augmentation from annotated training sets or direct question generation based on relevant knowledge points and documents, have expanded datasets but face challenges in mastering the inner logic of the problem during generation and ensuring the verifiability of the solutions. To address these issues, we propose RV-Syn, a novel Rational and Verifiable mathematical Synthesis approach. RV-Syn constructs a structured mathematical operation function library based on initial seed problems and generates computational graphs as solutions by combining Python-formatted functions from this library. These graphs are then back-translated into complex problems. Based on the constructed computation gr...
