TY  - RPRT
TI  - ER-JEPA: Experience Replay Improves Joint-Embedding Predictive Learning in Language Models
AU  - Jingnan Pu
AU  - Zi-En Fan
AU  - Feng Lian
PY  - 2026
UR  - https://arxiv.org/abs/2609.36952
ID  - 2609.36952
ER  -