Alex-GDA last-iterate convergence conjecture for convex-concave objectives

Let $f$ be a convex-concave objective function with $(L_x,L_y,L_{xy})$-Lipschitz gradients, and let Alex-GDA denote the extrapolated alternating gradient descent-ascent method. Alex-GDA last-iterate convergence conjecture. Alex-GDA exhibits last-iterate convergence to a Nash equilibrium of $f$. The conjecture extends the paper's bilinear convergence findings to general convex-concave objectives; whether this last-iterate convergence holds in that generality is left as future work.