Towards Self-Evolving Agentic Literature Retrieval
Summary
PaSaMaster is a self-evolving agentic literature retrieval system that iteratively refines search intent and produces evidence-grounded paper rankings, outperforming GPT-5.2 by 30% at 1% cost with zero hallucinations.
View Cached Full Text
Cached at: 05/15/26, 04:24 AM
Paper page - Towards Self-Evolving Agentic Literature Retrieval
Source: https://huggingface.co/papers/2605.14306 Authors:
,
,
,
,
,
,
,
,
,
Abstract
PaSaMaster is a self-evolving agentic literature retrieval system that improves academic search accuracy and cost efficiency through iterative intent analysis and evidence-based ranking.
Aslarge language modelsreshape scientific research,literature retrievalfaces a twofold challenge: ensuring source authenticity while maintaining a deep comprehension of academic search intents. While reliable, traditional keyword-centric search fails to capture complex research intents. Frontier LLMs can handle complex research intents, but their high cost and tendency to hallucinate remain key limitations. Here we introduce PaSaMaster, a self-evolving agenticliterature retrievalsystem that produces relevance-scored paper rankings withevidence-grounded recommendationsthrough iterativeintent analysis, retrieval, and ranking. It is built on three key designs. First, it transformsliterature retrievalfrom a one shot query--document matching problem into asearch processthat evolves over time, using ranked evidence to reveal gaps, refine intents, and guide follow-up searches. Second, it prevents hallucinated sources by treating retrieval as intent--paperrelevance rankingrather than generation. Finally, PaSaMaster improves cost efficiency by separating planning from retrieval: a frontier LLM is used only for intent understanding, while large scale retrieval and relevance scoring are delegated to customized corpora and lightweight models. Evaluated on thePaSaMaster Benchmarkacross 38 scientific disciplines, our system exposes the severe inaccuracy and incompleteness of traditional keyword retrieval (improving F1-score by 15.6X) and the unreliability of generative LLMs (which exhibithallucinationrates up to 37.79%). Remarkably, PaSaMaster outperforms GPT-5.2 by 30.0% at a mere 1% of thecomputational costwhile ensuring zero sourcehallucination: https://github.com/sjtu-sai-agents/PaSaMaster
View arXiv pageView PDFGitHub1Add to collection
Get this paper in your agent:
hf papers read 2605\.14306
Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash
Models citing this paper0
No model linking this paper
Cite arxiv.org/abs/2605.14306 in a model README.md to link it from this page.
Datasets citing this paper0
No dataset linking this paper
Cite arxiv.org/abs/2605.14306 in a dataset README.md to link it from this page.
Spaces citing this paper0
No Space linking this paper
Cite arxiv.org/abs/2605.14306 in a Space README.md to link it from this page.
Collections including this paper0
No Collection including this paper
Add this paper to acollectionto link it from this page.
Similar Articles
Rethinking Agentic Search with Pi-Serini: Is Lexical Retrieval Sufficient?
This paper introduces Pi-Serini, a BM25-based agentic search system that demonstrates lexical retrieval can suffice for deep search when agents refine queries, achieving high accuracy and reducing costs compared to default settings.
Multi-Turn Agentic Scientific Literature Search via Workflow Induction
This paper introduces PaperPilot, a multi-turn literature search agent that constructs executable DAGs of search operators and refines workflows through user feedback, achieving significant improvements in retrieval metrics over baseline models.
SAAS: Self-Aware Reinforcement Learning for Over-Search Mitigation in Agentic Search
SAAS introduces a reinforcement learning framework that enhances agent self-awareness to reduce unnecessary searches in LLM-based question answering systems, balancing accuracy and computational cost.
Consensus accelerates research with GPT-5 and Responses API
Consensus, a research assistant with 8 million users, has launched Scholar Agent—a multi-agent system built on GPT-5 and OpenAI's Responses API—that can synthesize peer-reviewed literature across 220 million papers in minutes. The system uses coordinated Planning, Search, Reading, and Analysis agents to mirror how human researchers work, reducing hallucinations and improving reliability over previous approaches.
SEMA-RAG: A Self-Evolving Multi-Agent Retrieval-Augmented Generation Framework for Medical Reasoning
SEMA-RAG is a self-evolving multi-agent RAG framework for medical question answering that decouples interpretation, exploration, and adjudication into three specialist agents, achieving significant accuracy improvements over baselines across multiple benchmarks.