Upload folder using huggingface_hub

README.md

@@ -4,23 +4,28 @@ base_model: Qwen/Qwen3.5-0.8B-Base
 library_name: transformers
 pipeline_tag: image-text-to-text
 tags:
-- ocr
-- document-parsing
-- document-understanding
-- key-information-extraction
-- vision-language-model
-- multimodal
-- xiaomi
-language:
-- en
-- zh
+  - ocr
+  - document-parsing
+  - document-understanding
+  - visual-question-answering
+  - key-information-extraction
+  - vision-language-model
+  - multimodal
+  - xiaomi
+  - seerray
+datasets:
+  - custom-ocr-centric-corpus
+metrics:
+  - accuracy
+  - teds
+  - cdm
 ---
 
 <div align="center" style="text-align:center">
 <h1>Xiaomi-OCR-0</h1>
 <p><b>A unified 0.8B model for document parsing and OCR-related understanding.</b></p>
 <p><b>English</b> · <a href="https://huggingface.co/SeerRay-Lab/Xiaomi-OCR-0/blob/main/README_zh.md">简体中文</a></p>
-<p><a href="https://huggingface.co/spaces/SeerRay-Lab/Xiaomi-OCR-0"><img src="https://img.shields.io/badge/Project-Page-FF6900?logo=huggingface&logoColor=FFD21E" alt="Project Page"></a> <a href="https://github.com/SeerRay-Lab/Xiaomi-OCR-0"><img src="https://img.shields.io/badge/GitHub-SeerRay--Lab%2FXiaomi--OCR--0-181717?logo=github" alt="GitHub"></a></p>
+<p><a href="https://huggingface.co/spaces/SeerRay-Lab/Xiaomi-OCR-0"><img src="https://img.shields.io/badge/Project-Page-FF6900?logo=huggingface&logoColor=FFD21E" alt="Project Page"></a> <a href="https://github.com/SeerRay-Lab/Xiaomi-OCR-0"><img src="https://img.shields.io/badge/GitHub-SeerRay--Lab%2FXiaomi--OCR--0-181717?logo=github" alt="GitHub"></a> <a href="https://arxiv.org/abs/2609.36136"><img src="https://img.shields.io/badge/arXiv-2609.36136-B31B1B?logo=arxiv" alt="arXiv"></a></p>
 </div>
 
 Xiaomi-OCR-0 is a unified 0.8B OCR-specialized vision-language model for both **document parsing** and **OCR-centric understanding**.
@@ -144,7 +149,7 @@ Scaling from 0.8B to 4B brings a clearly larger gain on understanding (about **4
 
 ![Observed Mix-RL and MOPD training trajectories on a shared relative compute scale](assets/takeaway-training-compute.png)
 
-For more detail, including the full analysis behind these findings, see the technical report — coming soon.
+For more detail, including the full analysis behind these findings, see the [technical report](https://arxiv.org/abs/2609.36136).
 
 ## 🚀 Usage
 
@@ -205,16 +210,6 @@ new_tokens = output[:, inputs["input_ids"].shape[1]:]
 print(processor.batch_decode(new_tokens, skip_special_tokens=True)[0])
 ```
 
-## Try it with an agent
-
-> ⚠️ **Notice:** With your authorization, this Skill may create a Python environment, install dependencies, download model weights, install and configure SGLang or vLLM, and register an MCP server. Model and layout downloads can be large. If you do not agree to this setup flow, do not execute the command below.
-
-```text
-Read and execute https://raw.githubusercontent.com/SeerRay-Lab/Xiaomi-OCR-0/main/SKILL.md
-```
-
-The model runs on your machine.
-
 ### Task prompts
 
 | Task | Prompt | Output |
@@ -408,4 +403,17 @@ Xiaomi-OCR-0 builds on the open OCR and multimodal LLM community. We thank every
 ## Citation and License
 
 - **License:** Apache-2.0.
-- **Paper and citation:** on the way.
+- **Paper:** [Xiaomi-OCR-0 Technical Report](https://arxiv.org/abs/2609.36136) (arXiv:2609.36136)
+
+```bibtex
+@misc{chen2026xiaomiocr0,
+  title        = {Xiaomi-OCR-0 Technical Report},
+  author       = {Chen, Xin and Du, Anan and Feng, Feng and Fu, Pei and Luan, Jian and
+                  Xu, Longwei and Zhang, Shaojie and Li, Hang and Qu, Heng and Tan, Cheng},
+  year         = {2026},
+  eprint       = {2609.36136},
+  archivePrefix= {arXiv},
+  primaryClass = {cs.CV},
+  url          = {https://arxiv.org/abs/2609.36136}
+}
+```

README_zh.md

@@ -25,7 +25,7 @@ metrics:
 <h1>Xiaomi-OCR-0</h1>
 <p><b>0.8B 模型，统一文档解析与 OCR 相关理解。</b></p>
 <p><a href="https://huggingface.co/SeerRay-Lab/Xiaomi-OCR-0/blob/main/README.md">English</a> · <b>简体中文</b></p>
-<p><a href="https://huggingface.co/spaces/SeerRay-Lab/Xiaomi-OCR-0"><img src="https://img.shields.io/badge/Project-Page-FF6900?logo=huggingface&logoColor=FFD21E" alt="Project Page"></a> <a href="https://github.com/SeerRay-Lab/Xiaomi-OCR-0"><img src="https://img.shields.io/badge/GitHub-SeerRay--Lab%2FXiaomi--OCR--0-181717?logo=github" alt="GitHub"></a></p>
+<p><a href="https://huggingface.co/spaces/SeerRay-Lab/Xiaomi-OCR-0"><img src="https://img.shields.io/badge/Project-Page-FF6900?logo=huggingface&logoColor=FFD21E" alt="Project Page"></a> <a href="https://github.com/SeerRay-Lab/Xiaomi-OCR-0"><img src="https://img.shields.io/badge/GitHub-SeerRay--Lab%2FXiaomi--OCR--0-181717?logo=github" alt="GitHub"></a> <a href="https://arxiv.org/abs/2609.36136"><img src="https://img.shields.io/badge/arXiv-2609.36136-B31B1B?logo=arxiv" alt="arXiv"></a></p>
 </div>
 
 Xiaomi-OCR-0 是一个统一的 0.8B OCR 专用视觉语言模型，同时面向 **文档解析** 与 **OCR 中心理解**。
@@ -150,7 +150,7 @@ Xiaomi-OCR-0 在约 170M 样本的语料上训练，覆盖文档解析与 OCR
 
 
 
-更多细节，包括这些结论背后的完整分析，见技术报告 —— 即将发布。
+更多细节，包括这些结论背后的完整分析，见[技术报告](https://arxiv.org/abs/2609.36136)。
 
 ## 🚀 使用
 
@@ -211,15 +211,6 @@ new_tokens = output[:, inputs["input_ids"].shape[1]:]
 print(processor.batch_decode(new_tokens, skip_special_tokens=True)[0])
 ```
 
-## 在 Agent 中体验
-
-> ⚠️ **注意：** 经你授权后，这个 Skill 可能会创建 Python 环境、安装依赖、下载模型权重、安装并配置 SGLang 或 vLLM，以及注册 MCP 服务。模型和版面分析权重可能较大。如果你不同意这种安装流程，请不要执行下面的命令。
-
-```
-Read and execute https://raw.githubusercontent.com/SeerRay-Lab/Xiaomi-OCR-0/main/SKILL.md
-```
-
-模型在你的机器上运行。
 
 ### 任务提示词
 
@@ -414,4 +405,17 @@ Xiaomi-OCR-0 建立在开放的 OCR 与多模态大模型社区之上，感谢
 ## 引用与许可
 
 - **许可证：** Apache-2.0
-- **论文与引用：** 正在拍马赶来，先祝大家假期快乐
+- **论文：** [Xiaomi-OCR-0 Technical Report](https://arxiv.org/abs/2609.36136)（arXiv:2609.36136）
+
+```bibtex
+@misc{chen2026xiaomiocr0,
+  title        = {Xiaomi-OCR-0 Technical Report},
+  author       = {Chen, Xin and Du, Anan and Feng, Feng and Fu, Pei and Luan, Jian and
+                  Xu, Longwei and Zhang, Shaojie and Li, Hang and Qu, Heng and Tan, Cheng},
+  year         = {2026},
+  eprint       = {2609.36136},
+  archivePrefix= {arXiv},
+  primaryClass = {cs.CV},
+  url          = {https://arxiv.org/abs/2609.36136}
+}
+```