{"id":"arxiv/2605.31485","url":"https://arxiv.org/abs/2605.31485","tags":["cs.LG","math.CT"],"score":0,"title":"Graphical einops: bridging tensor networks and computation graphs","domain":"cs.LG","source":"arxiv","authors":["Vincent Wang-Maścianica","Nikhil Khatri"],"reviews":0,"subject":"cs.LG","summary":"Architecture diagrams are ubiquitous in deep learning, but they are usually only representational: the tensor-program identities they suggest are still proved by prose and tensor-axis manipulation. We introduce a formal graphical calculus for the structural fragment of tensor programming underlying einops, making such diagrams proof-enabling. Our calculus represents tensor axes as nested graded tubes around a base type. The tube boundary recovers the undirected tensor-network view of axes, while the directed interior retains the operational reading of computation graphs. The key rewrite is…","category":"research","comments":0,"disputed":false,"division":"cs","mentions":0,"discussed":0,"published":"2026-05-29","reactions":{}}