Our paper on ideological generalisation in finetuned LLMs was accepted at NeurIPS 2026 – read it on arXiv
