We are glad to share with you that our recent conference paper “On Supervised Fine-tuning of Large Language Model for IELTS Writing Assessment System” from our MSAAI student Jinyang Li, which has been accepted in the 8th International Conference on Information Communication and Signal Processing (ICICSP 2025):
- Y. yang, R. T.-C. Hsung, T.-Y. Zhang, H. S.-H. Tsang, T. Y.L. Zhu, and W.-L. Lo, ” On Supervised Fine-tuning of Large Language Model for IELTS Writing Assessment System,” 2025 2nd International Conference on Innovation Management and Information Systems (ICICSP 2025), Xi’an, China, September, 2025.
Paper Abstract
Traditional IELTS writing tests are scored by human graders, which can be slow, expensive, and sometimes inconsistent. To address this, our team has developed an AI-powered system that uses advanced language models to grade IELTS essays automatically and provide detailed feedback.
We trained a large language model (LLM) on 5,000 high-quality IELTS essays so it can understand and score writing just like human experts. The system then gives users clear scores, helpful analysis, and specific advice on how to improve their writing.
Testing shows our AI grader scores essays more accurately and consistently than before, reducing errors by over 13% and increasing the accuracy rate from 45% to 60%. This new tool makes IELTS writing assessment faster, fairer, and more accessible—helping students prepare better and more efficiently.
Some photos of the authors and papers
Input Area of the Proposed IELTS Assessment System: Writing Topic and Essay Content Fields.
Comparison of IELTS Writing Scores. (Our approach: “Fine-tuned Model”)
Performance Evaluation. (Our approach: “Fine-tuned Model”)
