Paper ID: 2405.06331
LMD3: Language Model Data Density Dependence
John Kirchenbauer, Garrett Honke, Gowthami Somepalli, Jonas Geiping, Daphne Ippolito, Katherine Lee, Tom Goldstein, David Andre
We develop a methodology for analyzing language model task performance at the individual example level based on training data density estimation. Experiments with paraphrasing as a controlled intervention on finetuning data demonstrate that increasing the support in the training distribution for specific test queries results in a measurable increase in density, which is also a significant predictor of the performance increase caused by the intervention. Experiments with pretraining data demonstrate that we can explain a significant fraction of the variance in model perplexity via density measurements. We conclude that our framework can provide statistical evidence of the dependence of a target model's predictions on subsets of its training data, and can more generally be used to characterize the support (or lack thereof) in the training data for a given test task.
Submitted: May 10, 2024