DeconfoundLM
Official repository of the paper "Aligning Language Models with Observational Data: Opportunities and Risks from a Causal Perspective". The project explores methods to fine-tune large language models (LLMs) using observational data, tackling challenges like spurious correlations and confounding.