Abelio reframes a research goal before answering it: two physical gut-barrier repairs were archived, then tested
Abelio is an AI system that rewrites a research objective into several scientific perspectives, then pulls in elements from outside the usual discipline to build hypotheses. On 50 expert-curated objectives, its hypothesis-generation core beat four frontier large language models on scientific soundness and hypothesis diversity, and exceeded three of the four on distance from the literature. For intestinal barrier repair it proposed membrane-tethered PEG and cholesterol-anchored complementary DNA; both proposals were archived before experiments and then tested separately on damaged Caco-2 monolayers.