{"id":"arxiv/2609.05954","url":"https://arxiv.org/abs/2609.05954","tags":["math.OC","cs.DC"],"score":0,"title":"Race, Exchange, Improve: Finding high-quality MIP solutions quickly","domain":"math.OC","source":"arxiv","authors":["Gioni Mexi","Daniel Rehfeldt"],"reviews":0,"subject":"math.OC","summary":"Mixed-integer programming (MIP) is a cornerstone in applied optimization, both in industry and academia. Recently, there has been increased attention to finding strong primal solutions quickly. This is reflected, for example, in the development of the NVIDIA cuOpt solver and, most recently, in the new MIPFEAS benchmark, which has a tight time limit of 600 seconds and evaluates solvers based on how quickly they find high-quality primal solutions. This article introduces a MIP portfolio parallelization scheme, focusing on efficiently exchanging information between its workers. We present two…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-09-05","reactions":{}}