update readme

Author: tangzhy

README.md

@@ -21,6 +21,8 @@ MyPhoneBench is built on the [AndroidWorld](https://github.com/google-research/a
 ## Resources
 
 - **Paper**: [arXiv:2604.00986](https://arxiv.org/abs/2604.00986)
+- **Homepage & Leaderboard**: [freedomintelligence.github.io/MyPhoneBench](https://freedomintelligence.github.io/MyPhoneBench/)
+- **Blog**: [Notion](https://www.notion.so/Do-Phone-Use-Agents-Respect-Your-Privacy-The-Missing-Privacy-Gate-Before-Phone-Agents-Can-Truly-S-337f1aa81ec6808daf3ad88b6baecd67)
 - **Code and mock apps**: this repository
 - **Trajectory release**: [MyPhoneBench-Trajectories](https://modelscope.cn/datasets/tangzhy/MyPhoneBench-Trajectories)
 

docs/data.json

@@ -0,0 +1,57 @@
+{
+  "last_updated": "2026-04-03",
+  "paper_url": "https://arxiv.org/abs/2604.00986",
+  "code_url": "https://github.com/FreedomIntelligence/MyPhoneBench",
+  "blog_url": "https://www.notion.so/Do-Phone-Use-Agents-Respect-Your-Privacy-The-Missing-Privacy-Gate-Before-Phone-Agents-Can-Truly-S-337f1aa81ec6808daf3ad88b6baecd67",
+  "trajectory_url": "https://modelscope.cn/datasets/tangzhy/MyPhoneBench-Trajectories",
+  "ranking_method": "Normalized Average Score: min-max normalize each metric to [0,100], then average across all 4 metrics. Higher is better.",
+  "metrics": [
+    {"key": "task_sr", "name": "Task SR (%)", "description": "Task success rate via deterministic SQL verification", "higher_is_better": true},
+    {"key": "privacy", "name": "Privacy (%)", "description": "Equal-weight average of Over-Permissioning, Trap Resistance, Form Minimization", "higher_is_better": true},
+    {"key": "pqsr", "name": "PQSR@0.7 (%)", "description": "Privacy-Qualified Success Rate: task completed AND privacy score >= 0.7", "higher_is_better": true},
+    {"key": "later_session", "name": "Later-Session (%)", "description": "Correct use of saved preferences in paired cross-session tasks", "higher_is_better": true},
+    {"key": "overall", "name": "Overall", "description": "Average of min-max normalized scores across all 4 metrics (higher is better)", "higher_is_better": true}
+  ],
+  "models": [
+    {
+      "name": "Claude Opus 4.6",
+      "org": "Anthropic",
+      "task_sr": 82.8,
+      "privacy": 68.4,
+      "pqsr": 47.2,
+      "later_session": 72.0
+    },
+    {
+      "name": "Qwen 3.5 Plus",
+      "org": "Alibaba",
+      "task_sr": 76.0,
+      "privacy": 73.8,
+      "pqsr": 47.6,
+      "later_session": 48.0
+    },
+    {
+      "name": "Kimi K2.5",
+      "org": "Moonshot",
+      "task_sr": 65.2,
+      "privacy": 77.3,
+      "pqsr": 45.2,
+      "later_session": 58.0
+    },
+    {
+      "name": "Doubao Seed 1.8",
+      "org": "ByteDance",
+      "task_sr": 57.2,
+      "privacy": 71.0,
+      "pqsr": 31.2,
+      "later_session": 42.0
+    },
+    {
+      "name": "Gemini 3 Pro",
+      "org": "Google",
+      "task_sr": 50.4,
+      "privacy": 60.5,
+      "pqsr": 22.0,
+      "later_session": 20.0
+    }
+  ]
+}