TY  - RPRT
TI  - Separating Representation from Reconstruction Enables Scalable Text Encoders
AU  - Megi Dervishi
AU  - Mathurin Videau
AU  - Yann LeCun
PY  - 2026
UR  - https://arxiv.org/abs/2607.04011
ID  - 2607.04011
ER  -