Our paper on the implicit bias of gradient descent, reparametrization, and NNLS got accepted at the AISTATS 2025.