Add metadata and link to paper/code

README.md

@@ -1,24 +1,25 @@
 ---
 license: apache-2.0
+library_name: transformers
+pipeline_tag: image-text-to-text
 ---
 
 <div align="center">
 
-<h1> R1-Router: Learning to Route Queries across Knowledge Bases for Step-wise Retrieval-Augmented Reasoning </h1>
+<h1> MoRE: Mixture-of-Retrieval Experts for Reasoning-Guided Multimodal Knowledge Exploitation </h1>
 
 
 <h5 align="center"> 
 
 <a href='https://arxiv.org/abs/2505.22095'><img src='https://img.shields.io/badge/Paper-Arxiv-red'></a>
-<a href='https://huggingface.co/hmhm1229/R1-Router'><img src='https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-Models-blue'>
-<a href='https://huggingface.co/hmhm1229/R1-Router-3B'><img src='https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-Models-blue'>
+<a href='https://huggingface.co/hmhm1229/R1-Router-3B'><img src='https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-Models-blue'></a>
+<a href='https://github.com/OpenBMB/R1-Router'><img src='https://img.shields.io/badge/GitHub-Code-black'></a>
 
-Chunyi Peng<sup>1,3</sup>,
+Chunyi Peng<sup>1</sup>,
 Zhipeng Xu<sup>1</sup>,
 Zhenghao Liu<sup>1</sup>,
 Yishan Li<sup>3</sup>,
 Yukun Yan<sup>2</sup>,
-Zhiyuan Liu<sup>2</sup>,
 Yu Gu<sup>1</sup>
 Minghe Yu<sup>1</sup>
 Ge Yu<sup>1</sup>
@@ -30,8 +31,11 @@ Maosong Sun<sup>2</sup>
 </h5>
 </div>
 
+**MoRE** (Mixture-of-Retrieval Experts) is a novel framework that enables Multimodal Large Language Models (MLLMs) to collaboratively interact with diverse retrieval experts for more effective knowledge exploitation. This repository contains the **3B** version of the R1-Router, which was trained using Stepwise Group Relative Policy Optimization (Step-GRPO) to dynamically coordinate heterogeneous experts.
+
 ## News
-8.22 We upload [R1-Router-3B](https://huggingface.co/hmhm1229/R1-Router-3B).
+* **2026.04.03**: Our work is accepted by SIGIR2026 🎉🎉🎉!
+* **2025.08.22**: We upload [MoRE-3B](https://huggingface.co/hmhm1229/R1-Router-3B).
 
 ## Environment
 For training, answer generation, and evaluation processes:
@@ -59,26 +63,9 @@ For the image corpus, you can directly download [M-BEIR](https://huggingface.co/
 
 For the table corpus, you can download, embed and index Open-WikiTable following the [repository](https://github.com/sean0042/Open_WikiTable), or you can download directly the one we have already preprocessed from [here](https://huggingface.co/hmhm1229/table-retriever). 
 
-## Retrievers Preparation
-For the Text-Image Retriever, you can directly download [UniIR](https://huggingface.co/TIGER-Lab/UniIR)
-
-For the Table Retriever, you can train it with the help of [repository](https://github.com/sean0042/Open_WikiTable), or you can download it directly from [here](https://huggingface.co/hmhm1229/table-retriever). 
-
-## Datasets
-We have prepared all the text datasets in `./datasets`, for images you need to download them from:
-- `InfoSeek:` InfoSeek images can be downloaded from [OVEN](https://github.com/open-vision-language/oven/tree/main/image_downloads)
-- `Dyn-VQA:` Dynamic VQA images can be downloaded from [DynVQA_en.202412](https://github.com/Alibaba-NLP/OmniSearch/blob/main/dataset/DynVQA_en/DynVQA_en.202412.jsonl)
-- `WebQA:` WebQA images can be downloaded from [Google Drive](https://drive.google.com/drive/folders/19ApkbD5w0I5sV1IeQ9EofJRyAjKnA7tb)
-
 ## Training
-If you do not want to train the model, you can download [R1-Router](https://huggingface.co/hmhm1229/R1-Router) and skip this section to [Evaluation](#evaluation)
-### Data Synthesis
-If you want to use the ready-to-use synthetic data directly, you can skip this section to [Step-GRPO Training](#step-grpo-training)
+If you do not want to train the model, you can use this checkpoint and skip to [Evaluation](#evaluation).
 
-First, we need to synthesis the data step by step:
-```bash
-bash src/data_synthesis/data_synthesis.sh
-```
 ### Step-GRPO Training
 Our training framework is based on [EasyR1](https://github.com/hiyouga/EasyR1), only you need to do is to download it and replace some files with the files in `./Easy-R1`.
 Then start training with the command:
@@ -86,6 +73,7 @@ Then start training with the command:
 conda activate router
 bash examples/run_qwen2_5_vl_7b_stepgrpo.sh
 ```
+
 ## Evaluation
 We provide the evaluation pipeline for the R1-Router:
 ```bash
@@ -106,19 +94,16 @@ Our work is built on the following codebases, and we are deeply grateful for the
 - [OmniSearch](https://github.com/Alibaba-NLP/OmniSearch)
 
 ## Citation
-We appreciate your citations if you find our paper related and useful to your research!
-```
-@article{peng2025r1,
-  title={Learning to Route Queries across Knowledge Bases for Step-wise Retrieval-Augmented Reasoning},
-  author={Peng, Chunyi and Xu, Zhipeng and Liu, Zhenghao and Li, Yishan and Yan, Yukun and Wang, Shuo and Liu, Zhiyuan and Gu, Yu and Yu, Minghe and Yu, Ge and Sun, Maosong},
-  year={2025}
-  url={https://arxiv.org/abs/2505.22095}, 
+We appreciate your citations if you find our paper relevant and useful to your research!
+```bibtex
+@article{peng2025mixture,
+  title={Mixture-of-Retrieval Experts for Reasoning-Guided Multimodal Knowledge Exploitation},
+  author={Peng, Chunyi and Xu, Zhipeng and Liu, Zhenghao and Li, Yishan and Yan, Yukun and Wang, Shuo and Gu, Yu and Yu, Minghe and Yu, Ge and Sun, Maosong},
+  year={2025},
+  url={https://arxiv.org/abs/2505.22095}
 }
 ```
 
 ## Contact Us
 If you have questions, suggestions, and bug reports, please email us, we will try our best to help you.
-```
-hm.cypeng@gmail.com
-```
-
+`hm.cypeng@gmail.com`