IntrAgent: LLM Agent for Literature-Based Information Retrieval
Researchers introduce IntrAgent, an LLM-based agent designed for fine-grained information retrieval from scientific literature. The system addresses the new IntraView task, which automates content-grounded retrieval in response to research queries. IntrAgent mimics human reading behavior through a two-stage pipeline: a Section Ranking stage uses structural knowledge to prioritize relevant sections, followed by an Iterative Reading stage that extracts and synthesizes key details into concise answers. The approach aims to support analytical decisions in scientific research by faithfully grounding retrieval in provided content.
Key facts
- IntrAgent is an LLM-based agent for information retrieval from literature.
- The new task is called IntraView (INformation reTRieval through literAture reVIEW).
- IntrAgent mimics human reading behavior for retrieval.
- It uses a two-stage pipeline: Section Ranking and Iterative Reading.
- Section Ranking prioritizes relevant sections using structural knowledge.
- Iterative Reading extracts details and synthesizes grounded answers.
- The work is published on arXiv with ID 2604.22861.
- The goal is to support analytical decisions in scientific research.
Entities
Institutions
- arXiv