embedding-condensation

This paper observes that token embeddings in small language models condense into a narrow cone-like subspace, a phenomenon termed embedding condensation, and proposes a dispersion loss to counteract it, improving generalization.