AI models show high accuracy in number theory tasks, verifying conjectures

A new research paper explores the application of AI in number theory, evaluating the Qwen2.5-Math-7B-Instruct large language model on algorithmic and computational tasks. The model demonstrated high accuracy, achieving at least 0.95 on a benchmark of thirty algorithmic problems and thirty computational questions when provided with optimal hints. Additionally, the paper used a LightGBM classifier to empirically verify a conjecture about Dirichlet characters, predicting the modulus with over 93.9% accuracy based on statistical features of their initial zeros. AI

IMPACT Demonstrates potential for LLMs in specialized mathematical domains and empirical verification of number theory conjectures.

RANK_REASON The cluster contains an academic paper detailing computational experiments using AI models for number theory tasks. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• Ali Saraeb
• arXiv
• Hugging Face
• LightGBM
• Math StackExchange
• Qwen2.5-Math-7B-Instruct

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →