Research map / Large language models / Retrieval and knowledge
Template-Theorems Graph Construction to Enhance Mathematical Reasoning Capabilities of LLM
Yarong Lan, Yajing Xu, Huajun Chen
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add template-theorems graph retrieval context vs few-shot CoT | GPT-4.1 · AMC10 | accuracy: 75.12 -> 77.88 | Improved |
| add template-theorems graph retrieval vs few-shot CoT | DeepSeek-R1 · AIME24 | accuracy: 72.3 -> 76.2 | Improved |
| add template-theorems graph retrieval vs few-shot CoT | GPT-4o · MATH | accuracy: 74.8 -> 76.4 | Improved |
| add template-theorems graph retrieval vs few-shot CoT | DeepSeek-R1 · MMLU-Pro-Math | accuracy: 94.23 -> 94.52 | No clear effect |
| add few-shot examples over 0-shot | GPT-4o · AIME24 | accuracy: 13.20 -> 13.20 | No clear effect |
| add few-shot CoT examples over 0-shot | DeepSeek-R1 · AMC10 | accuracy: 85.25 -> 84.79 | Hurt |
| add both template and theorem knowledge vs template only (ablation) | GPT-4.1 · AMC10 | accuracy: 75.58 -> 77.88 | Improved |
| add theorem knowledge to template only (ablation) | GPT-4.1 · MATH | accuracy: 90.4 -> 91.4 | Improved |
| add template knowledge to theorem only (ablation) | GPT-4.1 · AMC12 | accuracy: 86.23 -> 86.23 | No clear effect |
| add advanced generation stage (ablation) | template-theorems graph · out-domain benchmarks (GPT-4.1) | accuracy | Improved |
| add verification step (ablation) | template-theorems graph · four benchmarks (GPT-4.1) | accuracy | Improved |
| add theorem refinement step (ablation) | retrieval · four benchmarks (GPT-4.1) | accuracy | Improved |
Closest papers by meaning
- Learning Theorem Rationale for Improving the Mathematical Reasoning Capability of Large Language Models (AAAI 2025 · similarity 0.92)
- DAG-Math: Graph-Guided Mathematical Reasoning in LLMs (ICLR 2026 · similarity 0.89)
- RealMath: A Continuous Benchmark for Evaluating Language Models on Research-Level Mathematics (NeurIPS 2025 · similarity 0.88)
- Semantic Search over 9 Million Mathematical Theorems (NeurIPS 2026 · similarity 0.88)
- One Example Shown, Many Concepts Known! Counterexample-Driven Conceptual Reasoning in Mathematical LLMs (ICML 2025 · similarity 0.88)
- GSM-Symbolic: Understanding the Limitations of Mathematical Reasoning in Large Language Models (ICLR 2025 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
