D-CPT Law: Domain-specific Continual Pre-Training Scaling Law for Large Language Models.
Haoran QueJiaheng LiuGe ZhangChenchen ZhangXingwei QuYinghao MaFeiyu DuanZhiqi BaiJiakai WangYuanxing ZhangXu TanJie FuWenbo SuJiamang WangLin QuBo ZhengPublished in: CoRR (2024)
Keyphrases
- language model
- domain specific
- language modeling
- information retrieval
- probabilistic model
- statistical language models
- speech recognition
- n gram
- language modelling
- document retrieval
- query expansion
- training set
- smoothing methods
- translation model
- retrieval model
- test collection
- passage retrieval
- relevance model
- document ranking
- error rate
- ad hoc information retrieval
- language models for information retrieval