Studying Large Language Model Generalization with Influence Functions

Abstract: When trying to gain better visibility into a machine learning model in order to understand and mitigate the associated risks, a potentially valuable source of evidence is: which training examples most contribute to a given behavior? Influence functions aim to answer a counterfactual: how would the model's parameters (and hence its outputs) change if a given sequence were added to the training set? While influence functions have produced insights for small models, they are difficult to scale to large …

abstract aim behavior counterfactual evidence examples functions influence language language model large language large language model machine machine learning machine learning model model generalization parameters risks studying training visibility