Saved in:
| Main Authors: | , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2604.11188 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866917402945519616 |
|---|---|
| author | Yu, Zixiong Rao, Jun Chen, Guhan Tian, Songtao Li, Bohan Wei, Jiansheng Zhang, Min Meng, Xiaojun |
| author_facet | Yu, Zixiong Rao, Jun Chen, Guhan Tian, Songtao Li, Bohan Wei, Jiansheng Zhang, Min Meng, Xiaojun |
| contents | Synthesizing high-quality mathematical reasoning data without human priors remains a significant challenge. Current approaches typically rely on seed data mutation or simple prompt engineering, often suffering from mode collapse and limited logical complexity. This paper proposes a hierarchical synthesis framework that formulates data synthesis as an unsupervised optimization problem over a constraint graph followed by semantic instantiation, rather than treating it as a direct text generation task. We introduce a Legislator-Executor paradigm: The Legislator adversarially evolves structured generation blueprints encoding the constraints of the problem, while the Executor instantiates these specifications into diverse natural language scenarios. This decoupling of skeleton design from linguistic realization enables a prioritized focus on constructing complex and diverse logical structures, thereby guiding high-quality data synthesis. Experiments conducted on a total of 10 models across the Qwen, Llama, Mistral, and Gemma series demonstrate that our method achieves notable results: models fine-tuned on 1K synthesized samples outperform widely-used datasets of comparable scale (LIMO, s1K) across eight mathematical benchmarks, exhibiting superior out-of-distribution generalization. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2604_11188 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | MathAgent: Adversarial Evolution of Constraint Graphs for Mathematical Reasoning Data Synthesis Yu, Zixiong Rao, Jun Chen, Guhan Tian, Songtao Li, Bohan Wei, Jiansheng Zhang, Min Meng, Xiaojun Computation and Language Artificial Intelligence Synthesizing high-quality mathematical reasoning data without human priors remains a significant challenge. Current approaches typically rely on seed data mutation or simple prompt engineering, often suffering from mode collapse and limited logical complexity. This paper proposes a hierarchical synthesis framework that formulates data synthesis as an unsupervised optimization problem over a constraint graph followed by semantic instantiation, rather than treating it as a direct text generation task. We introduce a Legislator-Executor paradigm: The Legislator adversarially evolves structured generation blueprints encoding the constraints of the problem, while the Executor instantiates these specifications into diverse natural language scenarios. This decoupling of skeleton design from linguistic realization enables a prioritized focus on constructing complex and diverse logical structures, thereby guiding high-quality data synthesis. Experiments conducted on a total of 10 models across the Qwen, Llama, Mistral, and Gemma series demonstrate that our method achieves notable results: models fine-tuned on 1K synthesized samples outperform widely-used datasets of comparable scale (LIMO, s1K) across eight mathematical benchmarks, exhibiting superior out-of-distribution generalization. |
| title | MathAgent: Adversarial Evolution of Constraint Graphs for Mathematical Reasoning Data Synthesis |
| topic | Computation and Language Artificial Intelligence |
| url | https://arxiv.org/abs/2604.11188 |