Search papers, labs, and topics across Lattice.
School of Computer Science and Technology, Beijing Jiaotong University
3
0
7
0
Text-only foundation models can perform surprisingly well on complex 3D spatial reasoning tasks, rivaling multimodal models, when equipped with a structured spatial representation derived from 3D reconstruction.
Retrieval-augmented agents get a serious reasoning boost by explicitly evaluating their own retrieval quality at each step, leading to state-of-the-art performance on multi-hop question answering.
GNNs can be made more robust to distribution shifts by explicitly identifying and mitigating spurious correlations between node features and labels using a Hilbert-Schmidt Independence Criterion (HSIC) based regularization.
