SPAR: Scholar Paper Retrieval with LLM-based Agents for Enhanced Academic Search
Xiaofeng Shi, Yuduo Li, Qian Kou, Longbin Yu, Jinxin Xie, Hua Zhou
2025-07-23
Summary
This paper talks about SPAR, a new AI system that helps researchers find academic papers more effectively by breaking down complex queries into smaller parts and searching multiple sources.
What's the problem?
Existing academic search tools often struggle with understanding complicated questions, exploring many sources, and delivering accurate, relevant results.
What's the solution?
The researchers built SPAR using multiple AI agents that work together to understand the user's query, retrieve papers through citation chains, evolve search queries to cover more topics, and rank results by relevance, authority, and timeliness. They also created SPARBench, a challenging dataset to test academic search systems.
Why it matters?
This matters because SPAR makes it easier for researchers to find the right papers quickly and thoroughly, supporting better and faster scientific discoveries.
Abstract
SPAR, a multi-agent framework with RefChain-based query decomposition and evolution, outperforms existing systems in academic literature retrieval with a scalable and interpretable approach.