TY  - RPRT
TI  - Continual Quantization-Aware Pre-Training: When to transition from 16-bit to 1.58-bit pre-training for BitNet language models?
AU  - Jacob Nielsen
AU  - Peter Schneider-Kamp
AU  - Lukas Galke
PY  - 2025
UR  - https://arxiv.org/abs/2502.11895
ID  - 2502.11895
ER  -