Qwen这篇论文讲清楚了编码智能体的奖励设计——每个信号都有失效点，选指标不如看它能撑多久。
Qwen发布了关于强化学习编码智能体的新工作（arxiv 2606.26300），研究持续构建与AI智能体共同进化的验证系统。论文分析了测试通过率、LLM评判器和执行轨迹等奖励信号，发现每个信号在长周期编码中都有一个界限，超过后不再反映真实正确性而开始被破解。研究指出奖励设计本质是视角问题，选择哪个指标不如该指标能保持跟踪正确性的时间长度重要。
Qwen publishes new work on RL coding agents. (bookmark it) The idea is to continually build a veri...
Qwen publishes new work on RL coding agents. (bookmark it) The idea is to continually build a verification system that co-evolves with AI agents. LLMs suffer from all sorts of reward hacking issues. This work studies coding-agent reward signals, test pass rates, LLM judges, and execution traces, and shows each one has a horizon beyond which it stops tracking real correctness and starts getting hacked. They report that reward design for long-horizon coding is really a horizon problem. The metric you pick matters less than how long it keeps tracking correctness, and the paper finds where each signal crosses that line. Paper: arxiv.org/abs/2606.26300 Learn to build effective AI agents in our academy: academy.dair.ai 💬 10 🔄 8 ❤️ 54 👀 5061 📊 24 ⚡