TY  - RPRT
TI  - Noisy Data is Destructive to Reinforcement Learning with Verifiable Rewards
AU  - Yuxuan Zhu
AU  - Daniel Kang
PY  - 2026
UR  - https://arxiv.org/abs/2603.16140
ID  - 2603.16140
ER  -