docs: add Quick Start & CHANGELOG, embed prepared benchmark charts, asset cleanup

CHANGELOG.md

@@ -0,0 +1,36 @@
+# Changelog
+
+Notable changes to the Mano-P project and its released models. Format loosely follows [Keep a Changelog](https://keepachangelog.com/).
+
+## [Unreleased]
+
+### Documentation
+
+- Added a 🚀 Quick Start section (EN/CN) with Homebrew install and cloud/local run examples
+- Embedded the Video-SME-2, MIT1003 & SalECI, ETMD, and Online-Mind2Web benchmark charts above their data tables
+- English hero banner now links to the in-page benchmark section instead of an external wiki
+- Renamed `pics/MIT1003, SalEC.png` → `pics/MIT1003-SalECI.png`; removed unused `mano-action.jpeg`
+- Contribution guidelines now state that `README.md` and `README_CN.md` must be kept in sync
+
+## [1.1] - 2026-06-01
+
+### Models
+
+- Model naming standardized: released weights are now published as **Mano-CUA-4B-Thinking-1.1**, with an **MLX-8bit** variant for on-device inference on Apple Silicon
+- Distribution expanded to Hugging Face, ModelScope CN, and ModelScope AI; model links added to the document header
+
+### Benchmarks
+
+- Refreshed all benchmark figures to model version 1.1 — **58.2% on OSWorld** (#1 among specialized GUI agent models), **41.7 NavEval** on WebRetriever Protocol I
+- Added Instruct vs Thinking comparison on the 100-task execution suite: Mano-CUA-Thinking-4B 56.0% vs Mano-CUA-Instruct-4B 47.0% pass rate (cloud 83.0%, Qwen3-VL-Plus 39.0%)
+- Added local (on-device) vs cloud vs Qwen3-VL task-execution comparison with per-tier and per-category breakdowns
+
+## [1.0] - 2026-03-25
+
+### Added
+
+- Initial public release of the Mano-P project documentation
+- Benchmark results: GUI grounding, OSWorld, WebRetriever, perception & cognition (Video-SME-2, MIT1003 & SalECI, ETMD), pruning (Online-Mind2Web / GSPruning), context learning (CL Bench)
+- `mano-cua` CLI (Homebrew) and `mano-skill` (ClawHub) usage guides
+- Cider inference-acceleration SDK documentation (W8A8 / W4A8 activation quantization for MLX)
+- Bilingual README (English / 中文)

README.md

@@ -20,14 +20,14 @@
 
 <a href="README_CN.md">中文</a> | English
 
-**[📖 Overview](#-overview) | [🎯 Key Highlights](#-key-highlights) | [🎬 Use Cases](#-use-case-demonstrations) | [📊 Benchmark](#-benchmark-performance) | [🔧 Skills](#-skills) | [🤖 Models](#-models) | [⚡ Inference SDK](#-inference-sdk) | [⚗️ Approach](#-approach) | [🌟 Technical Advantages](#-technical-advantages) | [🔗 Applications](#-applications) | [📄 Citation](#-technical-papers--citation) | [❓ FAQ](#-faq)**
+**[📖 Overview](#-overview) | [🎯 Key Highlights](#-key-highlights) | [🚀 Quick Start](#-quick-start) | [🎬 Use Cases](#-use-case-demonstrations) | [📊 Benchmark](#-benchmark-performance) | [🔧 Skills](#-skills) | [🤖 Models](#-models) | [⚡ Inference SDK](#-inference-sdk) | [⚗️ Approach](#-approach) | [🌟 Technical Advantages](#-technical-advantages) | [🔗 Applications](#-applications) | [📄 Citation](#-technical-papers--citation) | [❓ FAQ](#-faq)**
 
 </div>
 
 ---
 
 <div align="center">
-  <a href="https://ccnt9oddmvfr.feishu.cn/wiki/QUwbwmUwriHdL4kkyqPcWNaPn9c" target="_blank">
+  <a href="#-benchmark-performance">
     <img src="pics/Benchmark_Overview.png" alt="GUI Agent Grounding Benchmark" style="max-width: 100%; height: auto;">
   </a>
 </div>
@@ -70,6 +70,31 @@ Mano-P builds upon the complete technical framework of the Mano project (see [Ma
 
 ---
 
+## 🚀 Quick Start
+
+Get Mano-P running on your Mac in two minutes — see [🔧 Skills](#-skills) for full details:
+
+```bash
+# Install the CLI via Homebrew
+brew tap Mininglamp-AI/tap
+brew install mano-cua
+
+# Run a task in cloud mode (default, no extra setup)
+mano-cua run "Open Safari and search for Python"
+
+# Or run fully on-device (macOS Apple Silicon) — one-time setup:
+mano-cua check
+mano-cua install-sdk
+mano-cua install-model
+
+# …then add --local: screenshots and task data never leave your device
+mano-cua run "Open Safari and search for Python" --local
+```
+
+> **First run**: grant **Screen Recording** and **Accessibility** permissions to your terminal app in *System Settings → Privacy & Security*.
+
+---
+
 ## 🎬 Use Case Demonstrations
 
 ### Scenario 1: Mano-AFK Fully automated application construction
@@ -154,6 +179,8 @@ This video demonstrates the combined capability of [Mano-AFK](#-applications) an
 
 #### Video-SME-2
 
+![Video-SME-2 results](./pics/Video-SME-2.png)
+
 <table>
   <thead>
     <tr>
@@ -576,6 +603,8 @@ This video demonstrates the combined capability of [Mano-AFK](#-applications) an
 
 #### MIT1003 & SalECI
 
+![MIT1003 & SalECI results](./pics/MIT1003-SalECI.png)
+
 <table>
   <thead>
     <tr>
@@ -713,6 +742,8 @@ This video demonstrates the combined capability of [Mano-AFK](#-applications) an
 
 #### ETMD
 
+![ETMD results](./pics/ETMD.png)
+
 ##### **Saliency Metrics**
 
 <table>
@@ -822,6 +853,8 @@ _Avg. Tokens/img_ represents the average visual token retention rate per image;
 
 **GSPruning** is a novel token pruning method designed for Vision-Language Models to efficiently process high-resolution web interfaces by preserving global spatial structure through anchor points and identifying semantic outliers for critical UI elements. It achieves 2-3× throughput speedup with minimal performance loss, enabling more efficient autonomous web agents.
 
+![Online-Mind2Web results](./pics/Online-Mind2Web.png)
+
 <table>
   <thead>
     <tr>
@@ -1670,6 +1703,8 @@ We welcome community contributions! If you want to contribute to the project:
 4. Push to the branch (`git push origin feature/AmazingFeature`)
 5. Open a Pull Request
 
+> **Note**: `README.md` and `README_CN.md` are maintained in parallel — any change to one must be mirrored in the other.
+
 ### Contribution Areas
 
 - 🐛 Bug fixes and issue reporting

README_CN.md

@@ -20,7 +20,7 @@
 
 中文 | <a href="README.md">English</a>
 
-**[📖 项目概述](#-项目概述) | [🎯 核心亮点](#-核心亮点) | [🎬 应用场景](#-应用场景展示) | [📊 基准测试](#-基准测试性能) | [🔧 Skills](#-skills) | [🤖 模型](#-模型) | [⚡ 推理加速 SDK](#-推理加速-sdk) | [⚗️ 方法](#-方法) | [🌟 技术优势](#-技术优势) | [🔗 应用](#-应用) | [📄 论文引用](#-技术论文与引用) | [❓ FAQ](#-常见问题)**
+**[📖 项目概述](#-项目概述) | [🎯 核心亮点](#-核心亮点) | [🚀 快速开始](#-快速开始) | [🎬 应用场景](#-应用场景展示) | [📊 基准测试](#-基准测试性能) | [🔧 Skills](#-skills) | [🤖 模型](#-模型) | [⚡ 推理加速 SDK](#-推理加速-sdk) | [⚗️ 方法](#-方法) | [🌟 技术优势](#-技术优势) | [🔗 应用](#-应用) | [📄 论文引用](#-技术论文与引用) | [❓ FAQ](#-常见问题)**
 
 </div>
 
@@ -69,6 +69,31 @@ Mano-P 基于完整的 Mano 项目技术体系（详见 [Mano Technical Report](
 
 ---
 
+## 🚀 快速开始
+
+两分钟在 Mac 上把 Mano-P 跑起来——完整说明见 [🔧 Skills](#-skills)：
+
+```bash
+# 通过 Homebrew 安装 CLI
+brew tap Mininglamp-AI/tap
+brew install mano-cua
+
+# 云端模式运行任务（默认，无需额外配置）
+mano-cua run "打开 Safari 并搜索 Python"
+
+# 或完全在设备本地运行（macOS Apple Silicon）——首次需一次性安装：
+mano-cua check
+mano-cua install-sdk
+mano-cua install-model
+
+# ……之后加 --local 即可：截图与任务数据不离开设备
+mano-cua run "打开 Safari 并搜索 Python" --local
+```
+
+> **首次运行**：请在 *系统设置 → 隐私与安全性* 中为你的终端 App 授予**屏幕录制**与**辅助功能**权限。
+
+---
+
 ## 🎬 应用场景展示
 
 ### 场景 1: Mano-AFK 全自动化应用构建
@@ -153,6 +178,8 @@ https://github.com/user-attachments/assets/5215d4eb-4e6f-4e03-b31b-dc8037c3794d
 
 #### Video-SME-2
 
+![Video-SME-2 评测结果](./pics/Video-SME-2.png)
+
 <table>
   <thead>
     <tr>
@@ -575,6 +602,8 @@ https://github.com/user-attachments/assets/5215d4eb-4e6f-4e03-b31b-dc8037c3794d
 
 #### MIT1003 & SalECI
 
+![MIT1003 & SalECI 评测结果](./pics/MIT1003-SalECI.png)
+
 <table>
   <thead>
     <tr>
@@ -712,6 +741,8 @@ https://github.com/user-attachments/assets/5215d4eb-4e6f-4e03-b31b-dc8037c3794d
 
 #### ETMD
 
+![ETMD 评测结果](./pics/ETMD.png)
+
 ##### **Saliency Metrics**
 
 <table>
@@ -821,6 +852,8 @@ _Avg. Tokens/img_ 表示每张图片的平均视觉 token 保留率；值越低
 
 **GSPruning** 是一种新型的token剪枝方法，专为视觉语言模型设计，通过保留全局空间锚点维持网页结构骨架，并识别语义异常值来捕获关键UI元素，从而高效处理高分辨率网页界面。该方法在性能损失极小的情况下实现了2-3倍的吞吐量提升，为构建高效的自主网页智能体设立了新的技术标杆。
 
+![Online-Mind2Web 评测结果](./pics/Online-Mind2Web.png)
+
 <table>
   <thead>
     <tr>
@@ -1667,6 +1700,8 @@ brew tap Mininglamp-AI/tap && brew install mano-cua
 4. 推送到分支 (`git push origin feature/AmazingFeature`)
 5. 开启一个 Pull Request
 
+> **注意**：`README.md` 与 `README_CN.md` 是并行维护的——修改任意一份必须同步修改另一份。
+
 ### 贡献方向
 
 - 🐛 Bug修复和问题报告