GAN objective that minimizes the Wasserstein distance with a critic network constrained by Lipschitz conditions, improving training stability and mode coverage.