{"id":"arxiv/2602.06741","url":"https://arxiv.org/abs/2602.06741","tags":["cs.LG","physics.data-an","quant-ph"],"score":0,"title":"Disentanglement by means of action-induced representations","domain":"cs.LG","source":"arxiv","authors":["Gorka Muñoz-Gil","Hendrik Poulsen Nautrup","Arunava Majumder","Paulin de Schoulepnikoff","Florian Fürrutter","Marius Krumm","Hans J. Briegel"],"reviews":0,"subject":"cs.LG","summary":"Learning interpretable representations with variational autoencoders (VAEs) is a major goal of representation learning. The main challenge lies in obtaining disentangled representations, where each latent dimension corresponds to a distinct generative factor. This difficulty is fundamentally tied to the inability to perform nonlinear independent component analysis. Here, we introduce the framework of action-induced representations (AIRs) which models representations of physical systems given experiments (or actions) that can be performed on them. We show that, in this framework, we can…","category":"research","comments":0,"disputed":false,"division":"cs","mentions":0,"discussed":0,"published":"2026-02-06","reactions":{}}