Search papers, labs, and topics across Lattice.
Huazhong University of Science and Technology
2
0
4
Achieve state-of-the-art 3D scene understanding by dynamically adapting network parameters at test time, proving that input-aware adjustments can significantly boost performance with minimal overhead.
MLLMs can gain surprisingly strong 3D spatial reasoning abilities simply by tapping into the latent knowledge already present in video generation models.
