The Faithful Reasoning project focuses on combining the flexbility of LLMs and the rigorousness of Symbolic Reasoning, and generate faithful Neuro-symbolic reasoning for high-stakes use cases, such as determining whether a user has violated Terms of Service. Neuro-symbolic AI (also known as NeSy) is a field that combines the generalizability of Neural Networks and interpretability and control of Symbolic AI. The highly structured setup with potential capability for formal verification is a great fit for the legal field. Its generalizability is promising to lift up the burden of heavy legal document review and drafting
Instructed COT
Previous Publications
Wang, O. P., Bansal, T., Bai, R., Chui, E. M., & Gilpin, L. H. (2025). Follow My Lead: Logical Fallacy Classification with Knowledge-Augmented LLMs. arXiv preprint arXiv:2510.09970.