Lune

ACL Findings2024

MARIO: MAth Reasoning with code Interpreter Output - A Reproducible Pipeline

Minpeng Liao, Chengxi Li, Wei Luo, Jing Wu, Kai Fan

Abstract

Large language models (LLMs) have significantly improved in understanding natural language but still lack in mathematical reasoning, a hurdle on the path to true artificial general intelligence. The training of large language models, based on next-token prediction, struggles to capture the precise nature of mathematical reasoning, presenting both practical and theoretical challenges. In this paper, we address this challenge by enriching the data landscape and introducing a reasonable data format, enhanced the text analysis of the LLM with a capability to utilize a Python code interpreter. This dataset is derived from GSM8K and MATH and has been further refined through a combination of GPT annotations, human review, and selftraining processes. Additionally, we propose a tentative, easily replicable protocol for the fine-tuning of math-specific LLMs, which has led to a significant improvement in the performance of a 7B-parameter LLM on the GSM8K and MATH datasets. A solution generator and a value estimator are fine-tuned simultaneously in a multi-task fashion, while an outlier-free value model-based inference method is proposed to further boost the performance. We are committed to advancing the field of mathematical reasoning in LLMs. and, to that end, we will make the source code and checkpoints publicly available. 1

Ask Lune

Know this paper in minutes, not an afternoon.

Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.

22 top-tier papers have cited it since. Know what they found before your reviewers do.

Questions to start from

Lune can draw on

the full paper · the top-tier papers that cite it · related work

Get the answer

Free to start. No credit card required.

Lineage

  • ASyMOB: Algebraic Symbolic Mathematical Operations Benchmark cites Large Language Models are Zero-Shot Reasoners
  • Self-Refine: Iterative Refinement with Self-Feedback cites Chain-of-Thought Prompting Elicits Reasoning in Large Language Models
  • S²R: Teaching LLMs to Self-verify and Self-correct via Reinforcement Learning cites Let's Verify Step by Step
  • Large Language Models are Zero-Shot Reasoners cites Chain-of-Thought Prompting Elicits Reasoning in Large Language Models
  • Let's Verify Step by Step cites Large Language Models are Zero-Shot Reasoners
  • Let's Verify Step by Step cites Chain-of-Thought Prompting Elicits Reasoning in Large Language Models
  • Linking Process to Outcome: Conditional Reward Modeling for LLM Reasoning cites Let's Verify Step by Step
  • Linking Process to Outcome: Conditional Reward Modeling for LLM Reasoning cites Chain-of-Thought Prompting Elicits Reasoning in Large Language Models
  • Solver-Informed RL: Grounding Large Language Models for Authentic Optimization Modeling cites Let's Verify Step by Step
  • Solver-Informed RL: Grounding Large Language Models for Authentic Optimization Modeling cites Chain-of-Thought Prompting Elicits Reasoning in Large Language Models
Builds onCited byLeft to right by year; the more cited, the higher.

Cited by top-tier papers22

  1. ToolRL: Reward is All Tool Learning NeedsCheng Qian, Emre Can Acikgoz, Qi He, Hongru Wang et al.NeurIPS 2025 · 387 citations
  2. Solver-Informed RL: Grounding Large Language Models for Authentic Optimization ModelingYitian Chen, Jingfan Xia, Siyu Shao, Dongdong Ge et al.NeurIPS 2025 · 54 citations
  3. Linking Process to Outcome: Conditional Reward Modeling for LLM ReasoningZheng Zhang, Ziwei Shan, Kaitao Song, Yexin Li et al.ICLR 2026 · 16 citations
  4. S²R: Teaching LLMs to Self-verify and Self-correct via Reinforcement LearningRuotian Ma, Peisong Wang, Cheng Liu, Xingyan Liu et al.ACL 2025 · 13 citations
  5. ASyMOB: Algebraic Symbolic Mathematical Operations BenchmarkMichael Shalyt, Rotem Elimelech, Ido KaminerICML 2026 · 7 citations

Builds on22

  1. Chain-of-Thought Prompting Elicits Reasoning in Large Language ModelsJason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma et al.NeurIPS 2022 · 22,562 citations
  2. LoRA: Low-Rank Adaptation of Large Language ModelsEdward J. Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu et al.ICLR 2022 · 18,833 citations
  3. Large Language Models are Zero-Shot ReasonersTakeshi Kojima, Shixiang Shane Gu, Machel Reid, Yutaka Matsuo et al.NeurIPS 2022 · 8,168 citations
  4. Self-Refine: Iterative Refinement with Self-FeedbackAman Madaan, Niket Tandon, Prakhar Gupta, Skyler Hallinan et al.NeurIPS 2023 · 4,972 citations
  5. Let's Verify Step by StepHunter Lightman, Vineet Kosaraju, Yuri Burda, Harrison Edwards et al.ICLR 2024 · 3,045 citations

Related papers

  1. MathCoder: Seamless Code Integration in LLMs for Enhanced Mathematical ReasoningKe Wang, Houxing Ren, Aojun Zhou, Zimu Lu et al.ICLR 2024 · 188 citations
  2. Debate4MATH: Multi-Agent Debate for Fine-Grained Reasoning in MathShaowei Zhang, Deyi XiongACL Findings 2025 · 5 citations
  3. MuMath-Code: Combining Tool-Use Large Language Models with Multi-perspective Data Augmentation for Mathematical ReasoningShuo Yin, Weihao You, Zhilong Ji, Guoqiang Zhong et al.EMNLP 2024 · 3 citations
  4. MathCoder2: Better Math Reasoning from Continued Pretraining on Model-translated Mathematical CodeZimu Lu, Aojun Zhou, Ke Wang, Houxing Ren et al.ICLR 2025
  5. OpenMathInstruct-2: Accelerating AI for Math with Massive Open-Source Instruction DataShubham Toshniwal, Wei Du, Ivan Moshkov, Branislav Kisacanin et al.ICLR 2025
Dusk over the sea between two cliffs drawn in fine vertical lines