arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2605.24693cs.CL
收藏

CP-Agent: 一种用于反馈驱动竞赛编程的校准风险控制智能体

CP-Agent: A Calibrated Risk-Controlled Agent for Feedback-Driven Competitive Programming

The Hong Kong University of Science and Technology (Guangzhou)(香港科学与技术大学(广州))

Peisong Wang, Bowen Liu, Zehua Li, Yuyao Wang, Zhiwei Ma, Yuhan Li, Jia Li

AI总结:

提出CP-Agent,通过校准停止过程建模反馈驱动求解,结合双重粒度验证、测试增强和经验驱动自我进化机制,在不更新参数的情况下显著提升竞赛编程性能。

摘要:

大型语言模型在竞赛级编程中仍存在困难,而许多智能体解决方案依赖于大量的推理时采样或昂贵的多阶段后训练。我们研究了执行反馈何时能可靠地帮助LLM竞赛编程求解器,以及哪些机制支配着性能提升。我们将反馈驱动求解建模为校准停止过程,并识别出三个量:虚假接纳风险、针对不良程序的程序级证据以及活跃状态成功风险。在保留的轨迹校准和从预先声明的有限控制器清单中选择下,所得的结构性证书在虚假接纳之前为干净成功概率提供了下界。我们针对这些量实例化了机制:双重粒度验证、测试增强和经验驱动自我进化,从而得到CP-Agent。在不更新任何参数的情况下,CP-Agent在LiveCodeBench Pro上将Pass@1从25.8%提升至48.5%,并在ICPC-Eval上将Refine@5提高了11.0%。在三个LLM骨干网络上,CP-Agent处于成本-准确率效率前沿,消融实验表明每个组件主要影响其对应的证书量。

英文摘要:

Large language models still struggle with contest-level programming, while many agentic remedies rely on massive inference-time sampling or expensive multi-stage post-training. We study when execution feedback reliably helps an LLM CP solver and which mechanisms govern the gains. We model feedback-driven solving as a calibrated stopped process and identify three quantities: false-admission risk, program-level evidence against bad programs, and the active-state success hazard. Under held-out trace calibration and selection from a pre-declared finite controller manifest, the resulting structural certificate lower-bounds the clean success probability before false admission. We instantiate mechanisms targeting these quantities as Dual-Granularity Verification, Test Augmentation, and Experience-Driven Self-Evolving, yielding CP-Agent. Without updating any parameters, CP-Agent raises Pass@1 from 25.8\% to 48.5\% on LiveCodeBench Pro and improves Refine@5 by 11.0\% on ICPC-Eval. Across three LLM backbones, CP-Agent lies on the cost--accuracy efficiency frontier, and ablations show that each component primarily affects its corresponding certificate quantity.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑