SIGMA: Search-Augmented On-Demand Knowledge Integration for Agentic Mathematical Reasoning

SIGMA (Search-Augmented On-Demand Knowledge Integration for Agentic Mathematical Reasoning) is a new framework designed to enhance mathematical reasoning by utilizing specialized agents that independently reason and perform targeted searches, synthesizing results through a moderator mechanism. It achieves a 7.4% absolute performance improvement over existing systems on benchmarks like MATH500, AIME, and PhD-level science QA GPQA, highlighting its effectiveness in multi-agent knowledge integration for complex problem-solving. This framework is significant for practitioners as it offers a scalable method to improve reasoning accuracy and efficiency in AI applications.