metadata
base_model:
- intfloat/e5-base-v2
language:
- en
license: mit
pipeline_tag: text-retrieval
library_name: transformers
Agentic-R: Learning to Retrieve for Agentic Search
Introduction
This is the Agentic-R retriever model introduced in the paper Agentic-R: Learning to Retrieve for Agentic Search.
Agentic-R is a dense retriever specifically tailored for agentic search, where an agent interleaves multi-step reasoning with on-demand retrieval. Unlike retrievers designed for single-turn retrieval-augmented generation (RAG) that rely only on local passage utility, Agentic-R uses both local query-passage relevance and global answer correctness to measure passage utility in a multi-turn agentic search.
For detailed usage instructions, training scripts, and evaluation code, please refer to the 🧩 GitHub repository.
Citation
If you find this work helpful, please cite our paper:
@misc{liu2026agenticrlearningretrieveagentic,
title={Agentic-R: Learning to Retrieve for Agentic Search},
author={Wenhan Liu and Xinyu Ma and Yutao Zhu and Yuchen Li and Daiting Shi and Dawei Yin and Zhicheng Dou},
year={2026},
eprint={2601.11888},
archivePrefix={arXiv},
primaryClass={cs.IR},
url={https://arxiv.org/abs/2601.11888},
}