AI Insight
SIGMA is a new artificial intelligence framework that improves mathematical reasoning by using multiple specialized agents that independently search for knowledge, analyze problems from different perspectives, and synthesize information through a moderator mechanism. Each agent generates hypothetical passages to optimize retrieval for its specific analytical approach, making knowledge integration more efficient and context-aware. When tested on advanced mathematical benchmarks including MATH500, AIME, and PhD-level science questions, SIGMA achieved a 7.4% absolute performance improvement over existing open- and closed-source systems.
Why it matters
This multi-agent approach demonstrates a scalable method for tackling complex, knowledge-intensive problems that require both information retrieval and multi-step reasoning. The framework's success on PhD-level questions suggests potential applications in scientific research assistance, advanced education tools, and automated problem-solving systems across technical domains.
Understand the Science
Abstract: Solving mathematical reasoning problems requires not only accurate access to relevant knowledge but also careful, multi-step thinking. However, current retrieval-augmented models often rely on a single perspective, follow inflexible search strategies, and struggle to effectively combine information from multiple sources. We introduce SIGMA (Search-Augmented On-Demand Knowledge Integration for AGentic Mathematical reAsoning), a unified framework that orchestrates specialized agents to independently reason, perform targeted searches, and synthesize findings through a moderator mechanism. Each agent generates hypothetical passages to optimize retrieval for its analytic perspective, ensuring knowledge integration is both context-sensitive and computation-efficient. When evaluated on challenging benchmarks such as MATH500, AIME, and PhD-level science QA GPQA, SIGMA consistently outperforms both open- and closed-source systems, achieving an absolute performance improvement of 7.4%. Our results demonstrate that multi-agent, on-demand knowledge integration significantly enhances both reasoning accuracy and efficiency, offering a scalable approach for complex, knowledge-intensive problem-solving. We will release the code upon publication.
Source: SIGMA: Search-Augmented On-Demand Knowledge Integration for Agentic Mathematical Reasoning