Popular repositories Loading

1. BitDistiller BitDistiller Public

   [ACL 2024] A novel QAT with Self-Distillation framework to enhance ultra low-bit LLMs.

   Python 134 17

2. BitDecoding BitDecoding Public

   [HPCA 2026] A GPU-optimized system for efficient long-context LLMs decoding with low-bit KV cache.

   C++ 80 8

3. vlut.cpp vlut.cpp Public

   [MobiSys 2026] On-device parallel ultra-low-bit (ternary) LLM inference with LUT-based mpGeMM kernel.

   C++ 10 3