Search papers, labs, and topics across Lattice.
Yonsei University
1
0
GSRQ achieves a remarkable 22.20 percentage point accuracy gain on LongBench tasks, revolutionizing KV cache efficiency for LLMs at sub-1-bit quantization.
