Search papers, labs, and topics across Lattice.
6
0
7
7
Visual design no longer requires choosing between diffusion fidelity and code editability: orchestrating modular asset diffusion through a VLM-driven HTML/CSS coding loop delivers fully interactive, layer-decoupled graphic layouts.
Replacing standard residual connections with manifold-constrained hyper-connections boosts speaker recognition performance across multiple architectures.
City-scale 3D generation no longer stops at the front door: HoloWorld directly couples macro-level urban planning with micro-level, geometry-consistent interior generation in a single coherent pipeline.
WorldClaw can generate expansive, editable 3D worlds from text prompts while maintaining both global coherence and intricate local details.
Emotional inertia can significantly enhance emotion recognition accuracy in conversation, leading to a leap in performance over traditional methods.
Unleashing LLMs to "paint" with code lets you directly manipulate image generation, moving beyond black-box prompt engineering.
