Min-K%++: Improved Baseline for Detecting Pre-Training Data from Large Language Models.
Jingyang ZhangJingwei SunEric C. YeatsYang OuyangMartin KuoJianyi ZhangHao YangHai Helen LiPublished in: CoRR (2024)
Keyphrases
- language model
- training data
- language modeling
- n gram
- trec test collections
- word error rate
- document retrieval
- probabilistic model
- information retrieval
- speech recognition
- language modelling
- statistical language models
- query expansion
- ad hoc information retrieval
- retrieval model
- okapi bm
- smoothing methods
- context sensitive
- learning algorithm
- test collection
- pseudo relevance feedback
- training set
- decision trees
- query terms
- document ranking
- cross lingual
- language models for information retrieval
- translation model
- term dependencies
- labeled data
- information retrieval systems
- classification accuracy
- feature selection