PreprintarXiv.org2025
Bayesian Influence Functions for Hessian-Free Data Attribution
Philipp Alexander Kreer, Wilson Wu, Maxwell Adam, et al.
Proposes Bayesian influence functions (BIF) that replace Hessian inversion with loss landscape statistics from stochastic-gradient MCMC for scalable data attribution in large neural networks.
10Sep 30, 2025Neural Networks
arXiv