Chufei Luo
Hi! I'm a Ph.D. Student in Computer Engineering at Queen's University, Kingston, Canada. I'm jointly supervised by Xiaodan Zhu and Samuel Dahan.My main interest is in NLP model applications, and helping facilitate deployment of large language models in high-stakes domains. This includes addressing challenges such as interpretability, model robustness, and reasoning ability. Please find a selection of my projects below!
Legally Enforceable Hate Speech Detection for Public Forums
Accepted to the Findings of EMNLP 2023, Singapore. We propose to resolve issues of subjectivity in hate speech by drawing definitions from the law, and release a new dataset annotated by legal experts. Additionally, we present empirical results with the latest instruction-tuned models such as GPT-3.5, GPT-4, and Llama 2.
Prototype-based Interpretability for Legal Citation Prediction
Accepted to the Findings of ACL 2023, Toronto Canada. We enhance the interpretability of a language model by drawing parallels to a lawyer's thought process. A language model is trained to predict appropriate citations for input text based on latent space similarity to relevant legal reference, i.e. precedents and provisions of legislation. Our method produces a more inherently interpretable decision and achieves better performance compared to vanilla fine-tuning.
