UCL LLM Behaviour Control and Alignment

AI & ML interests

We focus on techniques to guide or influence the behaviour of a large language model (LLM), to produce more controlled or desired outputs. The main goals are: 1) Controlled Generation; 2) Prompt Tuning; 3) Model Alignment; 4) Response Conditioning; 5) Output Regulation; 6) Behavioural Fine-tuning.

Team members 1