TY  - RPRT
TI  - Recurrent Self-Improvement: Dynamic Cross-Loop On-Policy Distillation for Looped Language Models
AU  - Yi Wang
AU  - Rui Qian
AU  - Yu Li
AU  - Haoyang Yao
AU  - Wenjie Wang
PY  - 2026
UR  - https://arxiv.org/abs/2610.10623
ID  - 2610.10623
ER  -