Update update_paper_list.md

Author: boyugou

update_template_or_data/update_paper_list.md

@@ -1,3 +1,14 @@
+- [CUARewardBench: A Benchmark for Evaluating Reward Models on Computer-using Agent](https://arxiv.org/abs/2510.18596)
+    - Haojia Lin, Xiaoyu Tan, Yulei Qin, Zihan Xu, Yuchen Shi, Zongyi Li, Gang Li, Shaofei Cai, Siqi Cai, Chaoyou Fu, Ke Li, Xing Sun
+    - 🏛️ Institutions: Tencent Youtu Lab, PKU, NJU
+    - 📅 Date: October 21, 2025
+    - 📑 Publisher: arXiv
+    - 💻 Env: [Desktop]
+    - 🔑 Key: [benchmark], [dataset], [reward model], [computer-using agent], [CUARewardBench], [Unanimous Prompt Ensemble (UPE)]
+    - 📖 TLDR: This paper introduces CUARewardBench, the first benchmark for evaluating reward models tailored to computer-using agents. It includes step-level and trajectory-level annotations from tasks across 10 software types and 7 agent architectures. The study identifies the limitations of current reward models and proposes UPE, a prompting-based method that significantly improves evaluation accuracy.
+
+
+
 - [PolySkill: Learning Generalizable Skills Through Polymorphic Abstraction](https://arxiv.org/abs/2510.15863)
     - Simon Yu, Gang Li, Weiyan Shi, Peng Qi
     - 🏛️ Institutions: NEU, Uniphore