PaperPilot agent uses workflow induction for advanced scientific literature search

Researchers have developed PaperPilot, a novel multi-turn agent designed for scientific literature search. Unlike traditional agents that use fixed pipelines or simple language reasoning, PaperPilot frames search as workflow induction, creating an executable Directed Acyclic Graph (DAG) of search operators. This approach allows for greater control and refinement of search strategies through user feedback. Experiments show PaperPilot-9B significantly outperforms the base Qwen3.5-9B toolset agent in metrics like Hit@5, MRR, and nDCG@10, while also reducing execution errors. AI

IMPACT Introduces a novel agentic approach for scientific literature search, potentially improving research efficiency and discoverability.

RANK_REASON Research paper introducing a new agentic system for scientific literature search. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CL →

• alphaXiv
• arXivLabs
• CatalyzeX Code Finder for Papers
• Connected Papers
• CORE Recommender
• DagsHub
• Gotit.pub
• Hugging Face
• Influence Flower
• Litmaps
• PaperPilot
• Qwen3.5:9b
• ScienceCast
• scite Smart Citations

• paper
• product

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