Studying Large Language Model Generalization with Influence Functions

📖 Description

Studying Large Language Model Generalization with Influence Functions Roger Grosse?:, Juhan Bae?:, Cem Anil?: Nelson Elhage; Alex Tamkin, Amirhossein Tajdini, Benoit Steiner, Dustin Li, Esin Durmus, Ethan Perez, Evan Hubinger, Kamil? e Luko?i? ut? e, Karina Nguyen, Nicholas Joseph, Sam McCandlish Jared Kaplan, Samuel R. Bowman 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 language models (LLMs) due to the difficulty of computing an inverse-Hessian-vector product (IHVP). We use the Eigenvalue-corrected Kronecker-Factored App...

📊 Game Impacts

💭 Reactions

🔗 Sources

• [1] https://arxiv.org/abs/2308.03296