Search papers, labs, and topics across Lattice.
1
0
3
WorldCache achieves up to 3.7x speedups in diffusion-based world models by intelligently caching and selectively recomputing tokens, making interactive and long-horizon rollouts far more practical.