Pythia-14M
Emerging1papers using it
2026first seen
Pythia-14M is a dataset used to evaluate the susceptibility of language models by measuring their response to perturbations in the data distribution, resulting in the identification of interpretable clusters related to various patterns and structures.