Search papers, labs, and topics across Lattice.
Shanghai Jiao Tong University
2
0
4
InterSketch shows that interleaving visual sketches with textual reasoning, guided by self-correction and stepwise rewards, unlocks surprisingly strong long-horizon visual reasoning, even surpassing Gemini-3-Pro.
By focusing Mamba's attention on foreground voxels, Fore-Mamba3D achieves superior 3D object detection performance, overcoming limitations of previous Mamba-based methods that process the entire voxel sequence.