We have a new preprint online where we propose a self-adaptive, iterative framework for claim extraction. The goal is to adapt the way a claim is phrased using a preference signal coming from a fact checking model. Check it out here: arxiv.org/abs/2412.11653