Tag
This paper proposes a scalable supervised fine-tuning method for training language models to propose research hypotheses across disciplines. It has been accepted by ICML 2026 and the code is open source.