Matrix-vector multiplication implemented in off-the-shelf DRAM for Low-Bit LLMs Y Combinator 2025-05-04 23:35 Source Original site Comments