Search papers, labs, and topics across Lattice.
Shanghai Advanced Research Institute, Chinese Academy of Sciences, Shanghai, China
1
0
3
LVLMs hallucinate less when you intervene *before* they start generating, by cleaning up the initial Key-Value cache with modality-aware steering vectors.
