LivePlan：编程代理的在线监控与纠正引导
原标题：Online Monitoring and Corrective Steering of Programming Agents
AI 摘要
本文提出LivePlan框架，用于实时监控和纠正编程代理在解决GitHub问题时的行为低效和偏离。LivePlan将判断与建议解耦，使用确定性规则监控器检测问题，仅在检测到问题时才调用顾问LLM进行下一步纠正。在SWE-bench Verified和SWE-bench Pro上，LivePlan将问题解决率提升最高15.2%，平均提升9.9%，额外成本仅每实例0.08美元。
正文节选
Computer Science > Software Engineering Title:Online Monitoring and Corrective Steering of Programming Agents View PDF HTML (experimental) Abstract:Fixing GitHub issues in large-scale projects is a long-horizon task, especially when a fix requires changes across multiple locations or the issue description lacks the information needed to localize and repair it. As a result, agents traverse long trajectories that are prone to inefficiency and error: they drift away from their intended