Conditional Hypothesis Generation for LLM Text Analysis with Covariates

Arxiv CS.CL (NLP/language models)

The paper introduces a method for generating interpretable hypotheses from text using large language models, conditioned on researcher-specified covariates to discover language differences across outcomes like political affiliation. It targets computational social science applications.

llm

Arxiv CS.CL (NLP/language models)·Paiheng XuJing LiuWei Ai

8 days ago