Search papers, labs, and topics across Lattice.
The Hong Kong University of Science and Technology (
4
0
8
Cross-tokenizer On-Policy Distillation achieves superior efficiency and flexibility, enabling knowledge transfer between diverse model families without the constraints of shared tokenizers.
Autoregressive 3D layout generation can be both more physically plausible and significantly faster by repurposing existing 3D generative models.
Generate consistent stereo videos directly from RGB data, bypassing depth estimation and monocular-to-stereo conversion, with StereoWorld's novel camera-aware attention mechanisms.
MLLM training gets a 1.36x speed boost with Dynamic Hybrid Parallelism (DHP), which adaptively optimizes parallelism strategies to handle the data heterogeneity that plagues multimodal datasets.
