<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Projects on Juncheng Hu</title><link>https://jch.ai/projects/</link><description>Recent content in Projects on Juncheng Hu</description><generator>Hugo</generator><language>en-us</language><lastBuildDate>Wed, 06 Aug 2025 10:58:28 +0800</lastBuildDate><atom:link href="https://jch.ai/projects/index.xml" rel="self" type="application/rss+xml"/><item><title>Bottomup</title><link>https://jch.ai/projects/bottomup/</link><pubDate>Wed, 06 Aug 2025 10:58:28 +0800</pubDate><guid>https://jch.ai/projects/bottomup/</guid><description>&lt;h1 id="bottomupagent">BottomUpAgent&lt;/h1>
&lt;h2 id="rethinking-agent-design-from-top-down-workflows-to-bottom-up-skill-evolution">Rethinking Agent Design: From Top-Down Workflows to Bottom-Up Skill Evolution&lt;/h2>
&lt;div align="center">
&lt;p>&lt;a href="https://arxiv.org/abs/2505.17673" target="_blank" rel="noopener noreffer ">&lt;img
 class="lazyload"
 src="https://jch.ai/svg/loading.min.svg"
 data-src="https://img.shields.io/badge/arXiv-2505.17673-green.svg?style=flat-square&amp;amp;scale=1.5"
 data-srcset="https://img.shields.io/badge/arXiv-2505.17673-green.svg?style=flat-square&amp;amp;scale=1.5, https://img.shields.io/badge/arXiv-2505.17673-green.svg?style=flat-square&amp;amp;scale=1.5 1.5x, https://img.shields.io/badge/arXiv-2505.17673-green.svg?style=flat-square&amp;amp;scale=1.5 2x"
 data-sizes="auto"
 alt="https://img.shields.io/badge/arXiv-2505.17673-green.svg?style=flat-square&amp;amp;scale=1.5"
 title="arXiv" />&lt;/a> &lt;a href="https://www.python.org/" target="_blank" rel="noopener noreffer ">&lt;img
 class="lazyload"
 src="https://jch.ai/svg/loading.min.svg"
 data-src="https://img.shields.io/badge/Python-3.10-yellow.svg?style=flat-square"
 data-srcset="https://img.shields.io/badge/Python-3.10-yellow.svg?style=flat-square, https://img.shields.io/badge/Python-3.10-yellow.svg?style=flat-square 1.5x, https://img.shields.io/badge/Python-3.10-yellow.svg?style=flat-square 2x"
 data-sizes="auto"
 alt="https://img.shields.io/badge/Python-3.10-yellow.svg?style=flat-square"
 title="Python Version" />&lt;/a> &lt;a href="https://github.com/AngusDujw/Bottom-Up-Agent" target="_blank" rel="noopener noreffer ">&lt;img
 class="lazyload"
 src="https://jch.ai/svg/loading.min.svg"
 data-src="https://img.shields.io/github/stars/AngusDujw/Bottom-Up-Agent?style=flat-square&amp;amp;logo=github"
 data-srcset="https://img.shields.io/github/stars/AngusDujw/Bottom-Up-Agent?style=flat-square&amp;amp;logo=github, https://img.shields.io/github/stars/AngusDujw/Bottom-Up-Agent?style=flat-square&amp;amp;logo=github 1.5x, https://img.shields.io/github/stars/AngusDujw/Bottom-Up-Agent?style=flat-square&amp;amp;logo=github 2x"
 data-sizes="auto"
 alt="https://img.shields.io/github/stars/AngusDujw/Bottom-Up-Agent?style=flat-square&amp;amp;logo=github"
 title="GitHub stars" />&lt;/a>&lt;/p>
&lt;/div>
&lt;h2 id="introduction">Introduction&lt;/h2>
&lt;p>This repository contains the official implementation of the paper&lt;br>
&lt;strong>“Rethinking Agent Design: From Top-Down Workflows to Bottom-Up Skill Evolution”&lt;/strong>.&lt;br>
Our bottom-up agents learn skills through autonomous exploration and reasoning—starting from raw pixel inputs and simulated mouse/keyboard actions, evolving competence purely from experience.&lt;/p></description></item><item><title>DWA - Diversity-Driven Synthesis</title><link>https://jch.ai/projects/diversity-syn/</link><pubDate>Thu, 26 Sep 2024 00:00:00 +0000</pubDate><guid>https://jch.ai/projects/diversity-syn/</guid><description>&lt;h1 id="diversity-driven-synthesis-enhancing-dataset-distillation-through-directed-weight-adjustment">Diversity-Driven Synthesis: Enhancing Dataset Distillation through Directed Weight Adjustment&lt;/h1>
&lt;div align="center">
&lt;p>&lt;a href="https://arxiv.org/abs/2409.17612" target="_blank" rel="noopener noreffer ">&lt;img
 class="lazyload"
 src="https://jch.ai/svg/loading.min.svg"
 data-src="https://img.shields.io/badge/arXiv-2409.17612-green.svg?style=flat-square"
 data-srcset="https://img.shields.io/badge/arXiv-2409.17612-green.svg?style=flat-square, https://img.shields.io/badge/arXiv-2409.17612-green.svg?style=flat-square 1.5x, https://img.shields.io/badge/arXiv-2409.17612-green.svg?style=flat-square 2x"
 data-sizes="auto"
 alt="https://img.shields.io/badge/arXiv-2409.17612-green.svg?style=flat-square"
 title="arXiv" />&lt;/a> &lt;a href="" rel="">&lt;img
 class="lazyload"
 src="https://jch.ai/svg/loading.min.svg"
 data-src="https://img.shields.io/badge/NeurIPS-2024%20Spotlight-orange.svg?style=flat-square"
 data-srcset="https://img.shields.io/badge/NeurIPS-2024%20Spotlight-orange.svg?style=flat-square, https://img.shields.io/badge/NeurIPS-2024%20Spotlight-orange.svg?style=flat-square 1.5x, https://img.shields.io/badge/NeurIPS-2024%20Spotlight-orange.svg?style=flat-square 2x"
 data-sizes="auto"
 alt="https://img.shields.io/badge/NeurIPS-2024%20Spotlight-orange.svg?style=flat-square"
 title="NeurIPS" />&lt;/a> &lt;a href="https://github.com/AngusDujw/Diversity-Driven-Synthesis" target="_blank" rel="noopener noreffer ">&lt;img
 class="lazyload"
 src="https://jch.ai/svg/loading.min.svg"
 data-src="https://img.shields.io/github/stars/AngusDujw/Diversity-Driven-Synthesis?style=flat-square&amp;amp;logo=github"
 data-srcset="https://img.shields.io/github/stars/AngusDujw/Diversity-Driven-Synthesis?style=flat-square&amp;amp;logo=github, https://img.shields.io/github/stars/AngusDujw/Diversity-Driven-Synthesis?style=flat-square&amp;amp;logo=github 1.5x, https://img.shields.io/github/stars/AngusDujw/Diversity-Driven-Synthesis?style=flat-square&amp;amp;logo=github 2x"
 data-sizes="auto"
 alt="https://img.shields.io/github/stars/AngusDujw/Diversity-Driven-Synthesis?style=flat-square&amp;amp;logo=github"
 title="GitHub stars" />&lt;/a>&lt;/p>
&lt;/div>
&lt;h2 id="overview">Overview&lt;/h2>
&lt;p>Dataset distillation aims to compress a large dataset into a much smaller synthetic dataset while preserving its training utility. Our method, &lt;strong>Directed Weight Adjustment (DWA)&lt;/strong>, enhances dataset distillation by promoting diversity in the synthetic data through carefully directed adjustments to the optimization process.&lt;/p></description></item><item><title>Embedded ECG Diagnostic System with Data Communication</title><link>https://jch.ai/projects/ecg-fyp/</link><pubDate>Tue, 14 May 2024 00:00:00 +0000</pubDate><guid>https://jch.ai/projects/ecg-fyp/</guid><description>&lt;h1 id="embedded-ecg-diagnostic-system-with-data-communication">Embedded ECG Diagnostic System with Data Communication&lt;/h1>
&lt;p>&lt;strong>Undergraduate Final-Year Project&lt;/strong>&lt;br>
&lt;strong>NUS Research Institute&lt;/strong>&lt;br>
&lt;strong>Supervisor:&lt;/strong> &lt;a href="https://cde.nus.edu.sg/ece/staff/yung-c-liang/" target="_blank" rel="noopener noreffer ">Assoc. Prof. Yung C. Liang&lt;/a>&lt;br>
&lt;strong>Examiner:&lt;/strong> &lt;a href="https://cde.nus.edu.sg/ece/staff/wang-xinchao/" target="_blank" rel="noopener noreffer ">Assoc. Prof. Xinchao Wang&lt;/a>&lt;br>
&lt;strong>Oct 2023 - May 2024&lt;/strong>&lt;/p>
&lt;h2 id="overview">Overview&lt;/h2>
&lt;p>This project developed an embedded ECG diagnostic system for wearable, non-invasive body-signal recognition and data communication. The system combined ECG signal processing, hardware design, IoT communication, arrhythmia classification, cloud data streaming, and an interactive interface for real-time diagnostic feedback.&lt;/p></description></item><item><title>Django-Based SQL Online Judge System (System Optimization and Interface Design)</title><link>https://jch.ai/projects/sql-oj/</link><pubDate>Sat, 01 Jul 2023 00:00:00 +0000</pubDate><guid>https://jch.ai/projects/sql-oj/</guid><description>&lt;h1 id="django-based-sql-online-judge-system-system-optimization-and-interface-design">Django-Based SQL Online Judge System (System Optimization and Interface Design)&lt;/h1>
&lt;p>&lt;strong>Course Assessment Platform Development&lt;/strong>&lt;br>
&lt;strong>Oct 2022 - Jul 2023&lt;/strong>&lt;/p>
&lt;h2 id="overview">Overview&lt;/h2>
&lt;p>This project developed a Django-based SQL online judge system for automated course assessment and management. The system supported SQL language judging, question-bank management, exams, exercises, user accounts, and API-based notifications.&lt;/p>
&lt;h2 id="technical-scope">Technical Scope&lt;/h2>
&lt;ul>
&lt;li>Implemented database design and optimized system functions using ORM and MVC patterns.&lt;/li>
&lt;li>Built user account, SQL language judgment, and question-set recommendation systems.&lt;/li>
&lt;li>Designed and deployed question banks, exams, and exercises with API-based notifications.&lt;/li>
&lt;/ul>
&lt;h2 id="keywords">Keywords&lt;/h2>
&lt;p>Django, Nginx, Celery, Python, ORM, MVC, SQL Online Judge.&lt;/p></description></item></channel></rss>