TY  - RPRT
TI  - DynaCF: Mitigating Shortcut Learning in Reward Models via Dynamic Counterfactual Sensitivity
AU  - Fengyuan Liu
AU  - Yongliang Miao
AU  - Zirui He
AU  - Yanguang Liu
AU  - Fei Sun
AU  - Mengnan Du
PY  - 2026
UR  - https://arxiv.org/abs/2606.09043
ID  - 2606.09043
ER  -