Search papers, labs, and topics across Lattice.
Purdue University
1
0
3
Jointly optimizing visual token selection and LLM computation can drastically reduce inference costs while enhancing performance in multimodal tasks.
