Leak, Cheat, Repeat: Data Contamination and Evaluation Malpractices in Closed-Source LLMs.
Simone BalloccuPatrícia SchmidtováMateusz LangoOndrej DusekPublished in: EACL (1) (2024)
Keyphrases
- data processing
- data sets
- data collection
- statistical analysis
- data structure
- data analysis
- synthetic data
- data sources
- small number
- data mining techniques
- data quality
- experimental data
- domain experts
- input data
- database
- training data
- image data
- end users
- feature space
- missing data
- metadata
- raw data
- machine learning
- multiple sources
- multi source