AI RESEARCH

On the Invariants of Softmax Attention

arXiv CS.LG

ArXi:2605.02907v1 Announce Type: new Softmax attention maps every query--key interaction into a probability distribution, but the underlying structure remains largely unexplored. We define the \emph{energy field}, the row-centered attention logit, and show that it exhibits invariant properties across models, architectures, and inputs. Two classes of invariants emerge. \emph{Mechanism-level} invariants follow from the algebraic structure of softmax attention.