Search papers, labs, and topics across Lattice.
Honeycomb addresses the challenge of long-horizon video generation by introducing HexMemory, a constant-size scene memory representation that avoids quadratic memory growth. This approach utilizes a feed-forward writer to map new observations into six spatial and spatiotemporal planes, warping and fusing them with existing memory through confidence-weighted pooling and learned residual correction. Experiments on WorldScore and RealEstate10K show that Honeycomb achieves strong video generation quality and robust revisit consistency with fixed memory storage.
Video world models can now maintain persistent scene memory without ever-growing storage by warping and fusing low-rank features into a fixed-size HexMemory.
Video world models require persistent scene memory to maintain consistency during long-horizon video generation. Existing spatial memories accumulate RGB observations or latent features, increasing storage requirements as generation proceeds. We introduce Honeycomb, a video world model built on HexMemory, our proposed low-rank representation for storing scene features in a fixed-size memory with a total of six spatial and spatiotemporal planes. A feed-forward writer maps each generated chunk into new plane features. As the spatial coverage or temporal range expands, we warp the previous planes while preserving their dimensions, then fuse them with the new features through confidence-weighted pooling and a learned residual correction. A reader retrieves latents from HexMemory to condition subsequent video generation. The writer processes only observations from the new chunk, avoiding per-scene optimization and repeated processing of the full history. Experiments on WorldScore and RealEstate10K demonstrate strong video generation quality and robust revisit consistency while keeping HexMemory feature storage constant throughout generation. Code and additional visualizations are available on our project page at https://jackswl.github.io/honeycomb/.
