update readme

README.md

@@ -13,7 +13,7 @@ tasks:
 </p>
 
 <p align="center">
-        <a href="https://huggingface.co/TeleAI-AI-Flow/AI-Flow-Ruyi-7B-Preview0704">中文</a> &nbsp | &nbsp <a href="https://huggingface.co/TeleAI-AI-Flow/AI-Flow-Ruyi-7B-Preview0704/blob/main/README_en.md">English</a>
+        <a href="README.md">中文</a> &nbsp | &nbsp <a href="README_en.md">English</a>
         <br>
         🐱 <a href="https://github.com/TeleAI-AI-Flow/AI-Flow-Ruyi">GitHub</a> &nbsp&nbsp | &nbsp&nbsp 🤗 <a href="https://huggingface.co/TeleAI-AI-Flow/AI-Flow-Ruyi-7B-Preview0704">Hugging Face</a>&nbsp&nbsp | &nbsp&nbsp🤖 <a href="https://www.modelscope.cn/models/TeleAI-AI-Flow/AI-Flow-Ruyi-7B-Preview0704/">ModelScope</a>&nbsp&nbsp | &nbsp&nbsp 📑&nbsp <a href="https://www.arxiv.org/abs/2506.12479">Paper</a>
 </p>
@@ -23,6 +23,7 @@ tasks:
 
 ## 新闻
 
+* 🎉🎉[2025/7/4]：智传网（AI Flow）被全球资讯机构[Omdia](https://omdia.tech.informa.com/om137892/on-the-radar-teleai-brings-intelligence-to-the-network-edge-through-ai-flow)纳入短评，列为生成式 AI 落地应用的“重点观察”。
 * 🎉🎉[2025/7/4]：如意-7B预览版（AI-Flow-Ruyi-7B-Preview）发布
 
 ## 介绍
@@ -34,7 +35,7 @@ tasks:
 
 ## 如意-7B预览版
 
-如意-7B预览版（AI-Flow-Ruyi-7B-Preview）于7月4日发布。其最大参数量分支为7B，可分化出具有等效参数量为3B、4B、5B、6B的早退出分支。其中：
+为了让业界能亲身体验能够自由伸缩的“家族模型”，我们开源了如意-7B预览版（AI-Flow-Ruyi-7B-Preview），以展示我们在技术落地上的决心。如意-7B预览版（AI-Flow-Ruyi-7B-Preview）于7月4日发布。其最大参数量分支为7B，可分化出具有等效参数量为3B、4B、5B、6B的早退出分支。其中：
 * 3B、4B分支聚焦简单对话场景，其优势在于响应速度快、资源需求低；
 * 5B、6B分支则针对日常通用任务场景，在性能与响应速度之间寻求平衡；
 * 7B分支主要用于应对复杂问题，在多种能力维度上展现出较为全面的特性，但相对而言响应速度稍缓、资源需求略高。
@@ -209,3 +210,8 @@ print(generated_text)
       url={https://arxiv.org/abs/2506.12479}, 
 }
 ```
+
+## 致谢
+
+本工作受到学术界在“早退出机制”（Early-Exit Mechanism）方面长期探索的启发，特此向所有为该领域做出贡献的前辈学者致以诚挚谢意。限于篇幅与主题聚焦，我们仅在文献中引用了近期两篇具有代表性的研究 [[38](https://arxiv.org/abs/2404.16710),[39](https://ieeexplore.ieee.org/document/9414831)]，但这并不减损我们对其他先驱工作的敬意。期待与各位同仁一道，共同推动智能技术的普惠与平权。
+

README_en.md

@@ -13,16 +13,14 @@ tasks:
 </p>
 
 <p align="center">
-        <a href="https://huggingface.co/TeleAI-AI-Flow/AI-Flow-Ruyi-7B-Preview0704">中文</a> &nbsp | &nbsp <a href="https://huggingface.co/TeleAI-AI-Flow/AI-Flow-Ruyi-7B-Preview0704/blob/main/README_en.md">English</a>
+        <a href="README.md">中文</a> &nbsp | &nbsp <a href="README_en.md">English</a>
         <br>
         🐱 <a href="https://github.com/TeleAI-AI-Flow/AI-Flow-Ruyi">GitHub</a> &nbsp&nbsp | &nbsp&nbsp 🤗 <a href="https://huggingface.co/TeleAI-AI-Flow/AI-Flow-Ruyi-7B-Preview0704">Hugging Face</a>&nbsp&nbsp | &nbsp&nbsp🤖 <a href="https://www.modelscope.cn/models/TeleAI-AI-Flow/AI-Flow-Ruyi-7B-Preview0704/">ModelScope</a>&nbsp&nbsp | &nbsp&nbsp 📑&nbsp <a href="https://www.arxiv.org/abs/2506.12479">Paper</a>
 </p>
 
-#### Long long ago...
-> Deep within the Loong King's palace lay a divine rod, capable of infinite transformations, shrinking and growing at will. One day, finding himself at leisure, the Loong King mused to the rod: "With such formidable power, if only you could aid our Loong tribe in new endeavors." Suddenly, the rod spoke in reply: "I have an idea. What if this transformative ability were used to help humankind solve their problems?" No sooner said than done. The rod shimmered and transformed into an immensely powerful model, its "capabilities" freely scaling to match the complexity of any challenge. Beholding this marvel, the Loong King exclaimed, "Why, this is a true 'Ruyi' treasure—a wish-fulfilling aid to resolve all manner of troubles!" Thus, he named it "Ruyi" and sent it forth into the human world to offer its assistance.
-
 ## News
 
+* 🎉🎉[2025/7/4]：TeleAI’s AI Flow is now on the radar of global analyst firm [Omdia](https://omdia.tech.informa.com/om137892/on-the-radar-teleai-brings-intelligence-to-the-network-edge-through-ai-flow) as a generative-AI solution to watch.
 * 🎉🎉[2025/7/4]：AI-Flow-Ruyi-7B-Preview released!
 
 ## Introduction
@@ -34,7 +32,7 @@ tasks:
 
 ## AI-Flow-Ruyi-7B-Preview
 
-The AI-Flow-Ruyi-7B-Preview was released on July 4th. Its largest branch operates at 7B parameters and can differentiate into early-exit variants with **equivalent capacities of 3B, 4B, 5B, and 6B parameters**.
+To give the community a hands-on experience with a truly elastic “family of models,” we are open-sourcing the Ruyi-7B Preview (AI-Flow-Ruyi-7B-Preview), released on 4 July. Its largest branch contains 7 billion parameters and can spawn early-exit sub-networks with effective parameter counts of 3 B, 4 B, 5 B, and 6 B:
 
 Key branch specializations:
 * **3B/4B branches**: Optimized for simple dialogue scenarios, delivering **faster response times** with **minimal resource consumption**
@@ -47,8 +45,7 @@ Key branch specializations:
 |2|Layer 15|4B|AI-Flow-Ruyi-7B-E4B|Simple dialogue|
 |3|Layer 19|5B|AI-Flow-Ruyi-7B-E5B|Daily tasks|
 |4|Layer 23|6B|AI-Flow-Ruyi-7B-E6B|Daily tasks|
-|5|Layer 27|7B|AI-Flow-Ruyi-7B-E7B|Complex problems
-|
+|5|Layer 27|7B|AI-Flow-Ruyi-7B-E7B|Complex problems|
 
 ### Training process
 
@@ -211,4 +208,8 @@ print(generated_text)
       primaryClass={cs.AI},
       url={https://arxiv.org/abs/2506.12479}, 
 }
-```
+```
+
+## Acknowledgement
+
+We are deeply indebted to the academic community for the decades-long exploration of early-exit mechanisms, which inspired the development of the EESB approach. Owing to space constraints and thematic focus, we only cite two recent yet representative works [[38](https://arxiv.org/abs/2404.16710),[39](https://ieeexplore.ieee.org/document/9414831)]; nonetheless, our gratitude extends to every researcher who has advanced this field. We look forward to joining forces in making intelligence more inclusive and equitable for all.