{"id":"arxiv/2601.18005","url":"https://arxiv.org/abs/2601.18005","tags":["math.CO","cs.LG"],"score":0,"title":"Flow-based Extremal Mathematical Structure Discovery","domain":"math.CO","source":"arxiv","authors":["Gergely Bérczi","Baran Hashemi","Jonas Klüver"],"reviews":0,"subject":"math.CO","summary":"The discovery of extremal structures in mathematics requires navigating vast and nonconvex landscapes where analytical methods offer little guidance and brute-force search becomes intractable. We introduce FlowBoost, a closed-loop generative framework that learns to discover rare and extremal geometric structures by combining three components: (i) a geometry-aware conditional flow-matching model that learns to sample high-quality configurations, (ii) reward-guided policy optimization with action exploration that directly optimizes the generation process toward the objective while maintaining…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-01-25","reactions":{}}