OpenAI preprint, September 2026 · §4. A matrix-cone homomorphism with a full corner
4.4 Real tensor labels surviving rank-at-least-two maps:Finite-dimensional real vector spaces of dimension
yield arbitrarily many tensors in suitable tensor powers with linearly
independent images under tuples of such maps
Real tensors surviving maps of rank at least two.
Hover over a blue or green term in the text, or a node in the graph, to find it in the other. Select it to read its
declaration both in Lean and in words.
The Setting
Take a finite-dimensional real vector space with . Given an integer , real vector spaces , and -linear maps (), the tensor-product map sends the -fold real tensor power into the tensor product over . When every has rank at least , how many fixed tensors in can keep -linearly independent images under simultaneously for all choices of target spaces and maps, as grows?
The Claim
Theorem 1. Let be a finite-dimensional real vector space with , and let be an integer. There exist an integer and tensors such that for every -tuple of real vector spaces and every -tuple of -linear maps () each of rank at least , the family (1)✦
What Is Checked
The claim (1) is stated over a finite-dimensional real vector space with , an integer , and, for each integer , the real tensor power together with all -tuples of real vector spaces and -linear maps of rank at least via the induced maps .