Search papers, labs, and topics across Lattice.
Chung-Ang University, Seoul, Korea 11email:
1
0
2
Achieving 97.6% of full performance with just 5.6% of the visual tokens, AnchorPrune redefines efficiency in multimodal inference.
