TY  - RPRT
TI  - Large language models implicitly learn to straighten neural sentence trajectories to construct a predictive representation of natural language
AU  - Eghbal A. Hosseini
AU  - Evelina Fedorenko
PY  - 2023
UR  - https://arxiv.org/abs/2311.04930
ID  - 2311.04930
ER  -