Search papers, labs, and topics across Lattice.
2
0
3
0
AdaptiveEmbed reveals that tailoring representation capacity to individual sample needs can dramatically enhance multimodal retrieval performance.
Hallucinations in LVLMs can be cut by over 43% without sacrificing grounded object coverage, thanks to a novel verifier-guided approach.
