Weiqi Song

Undergraduate Student

Guangzhou,Guangdong

Overview

years of professional experience

Work History

Undergraduate Research Intern(remote)

Shanghai Zhangjiang Institute Of Mathematics

03.2024 - Current

1. Studying LLMs and LLM-based agent field.

2. Keeping up with cutting-edge research directions, exploring areas such as continual learning and knowledge editing(work in progress).

3. In practice scenario,I propose and validate improvements for the application of large language models in named entity recognition tasks, which significantly enhances LLMs performance in NER.

Group Member

Intelligent Control And Robot Research Group(GPNU)

09.2022 - 03.2024

In this period,I completed the basic scientific research training in this research group, and have the ability of independent scientific research.

Education

Bachelor of Engineering - Electronic Information Engineering

Guangdong Polytechnic Normal University

Guangzhou City, Guangdong Province

04.2001 -

Skills

Proficiency with python
Proficiency with Pytorch
Proficiency with Data Processing

Familiarity Distributed Training
Scientific Writing

Projects

1.Temperature recognition and detection

In this project, the neural network model is used to identify the thermal image transmitted by the microcontroller to the server. I was responsible for using yolov3 model to recognize and identify thermal images transmitted to server side(including training model).

2. 2D Human keypoint recoginition

In this project, the neural network model is used to identify and detect human joint keypoint, and transmit their coordinates to Unity3D for 3D human model. I was responsible for training a NN model that refers to HRNet.

3. LLMs in Named Entity Recoginition

In this project, the LLM is used to recognize and extract named entity. Due to mediocre performance in traditional methods for LLM, I propose a great improvement method for LLM in NER, significantly enhancing LLM's performance.I was responsible for all aspect.

Publication

ReplaceNER:A Efficient And Simple Method For Flat Named Entity Recognition Via Instruction Tuning

(waiting for AAAI-25 to be submitted)

Competitions

1.The first GPNU AI algorithm competition(third prize)

2. The first GPNU & Desay CV AI algorithm competition(third prize)

3. College students innovation and entrepreneurship training competition(school-level leader)

Timeline