BitNet Distillation
Research paper by Xun Wu and collaborators. Published 2025-10-15. Topics: BitNet Distillation, BitDistill, SubLN, multi-head attention distillation, continual pre-training.
Resource information
- Authors
- Xun Wu, Shaohan Huang, Wenhui Wang, Ting Song, Li Dong, Yan Xia, Furu Wei
- Research area
- Machine learning
- Application domain
- General AI research
- AI methods
- Large language models
- Research topics
- cs.LG, cs.CL
- arXiv record date
- 2025-10-15
- Publication date
- 2025-10-15
- Publication
- arXiv
- Upstream last modified
- Oct 17, 2025
- License
- CC BY 4.0
- Last record update
- Nov 3, 2026
Abstract
In this paper, we present BitNet Distillation (BitDistill), a lightweight pipeline that fine-tunes off-the-shelf full-precision LLMs (e.g., Qwen) into 1.58-bit precision (i.e., ternary weights -1, 0, 1) for specific downstream tasks, achieving strong task-specific performance with minimal computational cost. Specifically, BitDistill incorporates three key techniques: the SubLN module, as introduced in BitNet; multi-head attention distillation, based on MiniLM; and continual pre-training, which serves as a crucial warm-up step to mitigate the scalability issue of the performance gap between finetuned full-precision and 1.58-bit LLMs on specific tasks. Experimental results show that BitDistill achieves performance comparable to the full-precision counterpart models across model size, while enabling up to 10x memory savings and 2.65x faster inference on CPUs. Code is available at https://github.com/microsoft/BitNet.