Agents-K1: Towards Agent-native Knowledge Orchestration
2026-06-11 • Artificial Intelligence
Artificial Intelligence
AI summaryⓘ
The authors created Agents-K1, a new system that turns whole scientific papers into detailed knowledge graphs instead of just using abstracts or simple citation info. Their approach includes advanced parsing to identify key scientific elements like claims and evidence, a trained extraction model, and a tool that combines web search with graph-based data retrieval. They used this system to process millions of papers, producing a large scientific knowledge graph called Scholar-KG, which they partially released publicly. Their experiments show that Agents-K1 works well for extracting scientific information and supporting complex reasoning across multiple documents.
Large Language ModelsKnowledge GraphInformation ExtractionScientific ReasoningMultimodal ParsingCitationsMulti-hop ReasoningGraph RetrievalAgent OrchestrationScholarly Data
Authors
Zongsheng Cao, Bihao Zhan, Jinxin Shi, Jiong Wang, Fangchen Yu, Zhijie Zhong, Zijie Guo, Tianshuo Peng, Zhuo Liu, Yi Xie, Xiang Zhuang, Yue Fan, Runmin Ma, Shiyang Feng, Xiangchao Yan, Anran Liu, Peng Ye, Wenlong Zhang, Shufei Zhang, Chunfeng Song, Fenghua Ling, Jie Zhou, Liang He, Bo Zhang, Lei Bai
Abstract
Current LLM-based research agents have advanced through agent orchestration, yet largely overlook scientific knowledge orchestration. Existing works often reduce papers to abstracts, surface mentions, and flat \texttt{cites} edges, omitting key entities, claims, evidence, mechanisms, and method lineages essential for scientific reasoning. To this end, we introduce \textbf{Agents-K1}, an end-to-end knowledge orchestration pipeline that converts raw documents into agent-native scientific knowledge graphs. Agents-K1 integrates three components under a unifying theoretical foundation: a multimodal parser whose five-module schema captures entities, multimodal evidence, citations, and typed inter-entity relations across the full paper rather than abstracts alone; a 4B information-extraction backbone trained with GRPO under a rule-based reward; and a graphanything CLI, a tri-source agent interface that unifies web search, multimodal graph retrieval, and cross-document traversal. On top of this, we process 2.46 million scientific papers across six subjects to produce \textbf{Scholar-KG}, of which we release a one-million-paper subset, and the full Scholar-KG is accessible via the SCP link below. The same pipeline can be extended to general-domain corpora and to schema-conformant data synthesis. Extensive experiments demonstrate that Agents-K1 achieves superior performance in scientific information extraction, knowledge graph construction, and multi-hop scientific reasoning.