Preprint
Oct 2026
Latent-Foresight: End-to-End Learning Predictable Representations for Latent World Models
Extensive experiments show that the proposed Latent-Foresight approach learns more temporally coherent latent representations and consistently outperforms two-stage baselines across multiple future scene understanding tasks and prediction horizons, while eliminating separate training stages, including during high-resol...
Efstathios Karypidis, Spyros Gidaris, Nikos Komodakis
· 0 citations