1 paper · 1 filter
Greyson Brothers, Willa Mannering, Amber Tien +1
We explore the Iterative Inference Hypothesis (IIH) within the context of transformer-based language models, aiming to understand how a model's latent representations are progressi…