AI-Driven Learning and Regeneration of Analog Circuit Designs

Supervisor: Prof. Xiangyu Meng

Duration: August 2023 - August 2024

Developed an AI-based framework to learn and regenerate analog circuits from academic literature. Achieved an average accuracy of 97% in target detection within the circuit extractor module using deep learning techniques like Faster R-CNN.

Preprint: Link

FrGNet: A Fourier-Guided Weakly-Supervised Framework for Nuclei Instance Segmentation

Collaborator: Dr. Peng Ling

Duration: January 2024 - September 2024

Proposed a weakly-supervised deep learning framework for nuclei instance segmentation in histopathologic images. Introduced the Fourier Guidance Module and Guide-based Instance Level Contrastive Module to enhance segmentation performance.

Trustworthy Multi-Modal Benchmark for Medical Large Vision Language Models (Med-LVLMs)

Supervisor: Prof. Suhang Wang

Duration: June 2024 - Present

Developing a new benchmark for evaluating the trustworthiness of Med-LVLMs across dimensions like trustfulness, safety, robustness, fairness, and privacy.

Large Language Models for SVG Flowchart Generation

Supervisor: Prof. Sheng Li

Duration: July 2024 - Present

Developed a framework using Large Language Models (LLMs) to generate SVG flowcharts from textual descriptions. Collected and processed datasets of flowchart diagrams extracted from academic papers.

Automatic Labeling of Metal Blocks in GDS Layout Using Large Language Models

Supervisor: Prof. Xiangyu Meng

Duration: May 2024 - Present

Employed LLMs to automatically label metal blocks in GDS layouts based on positions and connections. Fine-tuned the Meta-Llama-3-8B model to achieve high labeling accuracy.

Machine Learning Based Dynamic Optimization Framework for Analog Circuit Sizing

Supervisor: Prof. Xiangyu Meng

Duration: June 2023 - March 2024

Developed a machine learning-based framework to dynamically optimize the sizing of analog circuits. Improved efficiency and performance in circuit design.

Understanding Positive Customer Engagement: The Impact of Cognition and Emotion on Behavior

Supervisor: Dr. Luning Zang

Duration: October 2023 - February 2024

Explored how cognitive and emotional factors influence positive customer engagement. Applied the BERT framework to achieve approximately 92% accuracy in multi-label text classification tasks.