Search papers, labs, and topics across Lattice.
Work done during internship at Tencent. 鈥orresponding authors
2
0
3
Early decision-making in multimodal reasoning can cut inference time without sacrificing accuracy, thanks to a novel dual evaluation of model competencies.
MeshWeaver achieves a groundbreaking 18% compression ratio while generating high-fidelity meshes with over 16K faces by weaving geometry directly into the autoregressive process.
