Harness^2:面向开放世界的递归智能体控制框架
Harness^2: Recursive Agent Harnessing for an Open World
AI总结:
Harness^2是面向开放世界的递归智能体控制框架,通过任务级与控制框架级递归解决开放世界下RSI的挑战,在多个基准上提升了GLM-5和Gemini 3.5 Flash的性能。
摘要:
大语言模型(LLM)智能体的能力不仅取决于其底层模型,还取决于其执行控制框架(harness)。递归自我改进(RSI)为自动演化这些控制框架提供了一种有前景的方法。然而,在开放世界场景中部署递归自我改进仍面临两个关键挑战:第一,开放世界任务通常在控制框架设计阶段未被见过,且在测试时是新颖的,没有真实值验证器来可靠地指导在线控制框架的编辑;第二,递归自我改进过程仍然高度依赖特定任务,无法泛化到新颖场景。我们提出Harness^2,这是一个即插即用的递归智能体控制框架,通过两个相互关联的递归层级解决这些挑战:任务级递归采用对比式“提出-探测-组合”步骤来优化当前任务的控制框架,该步骤无需真实值验证器即可可靠指导编辑,且可通过并行或顺序递归进一步扩展;控制框架级递归则从这些编辑中提取经验,以更新持续改进的控制框架,使智能体能更好地适应未来任务。在三个专业基准和四个模型-控制框架配置上,单步操作始终能提升GLM-5和Gemini 3.5 Flash的性能,相较于其基础控制框架,在WorkBuddy-Bench上的提升可达11.5分,在JobBench上可达4.5分,在LAB上可达11.8个百分点,且随着任务级递归预算的增加,性能会进一步提升。
英文摘要:
LLM agents' capabilities depend not only on their underlying models but also on their execution harnesses. Recursive self-improvement (RSI) offers a promising approach to automatically evolve these harnesses. However, deploying RSI in open-world settings still faces two critical challenges. First, open-world tasks are typically unseen during the harness design phase and novel at test time, leaving no ground-truth verifiers to reliably guide online harness edits. Second, the RSI process remains highly task-specific and fails to generalize to novel scenarios. We introduce Harness^2, a plug-and-play, recursive agent harnessing framework that addresses these challenges through two connected levels of recursion. Task-level recursion refines the current task's harness using a contrastive Propose-Probe-Compose step that reliably guides edits without ground-truth verifiers, and can be scaled further via parallel or sequential recursion. Harnessing-level recursion then extracts experience from these edits to update a continual improvement harness, enabling the agent to better adapt to future tasks. Across three professional benchmarks and four model-harness configurations, a single step consistently improves GLM-5 and Gemini 3.5 Flash over their base harnesses, yielding gains of up to 11.5 points on WorkBuddy-Bench, 4.5 points on JobBench, and 11.8 percentage points on LAB, with further gains as the task-level recursion budget increases.
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。