[Request] Need Arxiv endorser for grokking interpretability paper (ssrn accepted)
Hi, I'm an independent researcher submitting to Arxiv for the first time
and need an endorser in cs.LG or cs.AI.
The paper introduces Cycle Closure Count (CCC), a functional probe for
algebraic structure in grokking, and shows that apparent "quotient-first
learning" is a coordinate artifact.
the paper is accepted by SSRN at http://dx.doi.org/10.2139/ssrn.6888418
welcome feedback, and thanks a lot if could endorse for arxiv.