Tag
ReMoMask-2 improves text-to-motion generation by embedding retrieval directly into the generator's latent space, eliminating representation gaps and achieving state-of-the-art results on benchmarks like KIT-ML and SnapMoGen.