Do Localization Methods Actually Localize Memorized Data in LLMs? A Tale of Two Benchmarks.
Ting-Yun ChangJesse ThomasonRobin JiaPublished in: NAACL-HLT (2024)
Keyphrases
- data sets
- original data
- data mining techniques
- data analysis
- synthetic data
- statistical methods
- training data
- data quality
- noisy data
- missing values
- spectral clustering
- statistical analysis
- raw data
- high dimensional data
- complex data
- complex structures
- predictive model
- data processing
- knowledge discovery
- data structure
- small number
- data collection
- image data
- data points
- high quality
- incomplete data
- statistical tests
- data mining methods
- data mining applications
- database systems
- data reduction
- significant improvement
- computational cost
- data sources
- end users
- data mining algorithms