Search papers, labs, and topics across Lattice.
Hong Kong University of Science and Technology
1
0
Block-GTQ slashes quantization error by up to 80%, transforming how we optimize key-cache performance in long-context models.
