发表机构
Microsoft Research（微软研究院）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
本研究提出fork-and-flush干预，通过分叉平行轨迹并选择最高分继续，帮助自动研究智能体逃离想法盆地，在13项长时程任务中相对提升66.0%和44.4%。
AI 中文摘要
自动研究智能体通过反复提出候选解决方案、评估它们并利用反馈指导后续实验来处理开放性问题。我们表明，同一智能体在同一任务上的独立运行通常会在显著不同的分数上达到平台期，即使在投入大量额外计算后，这些差距仍然存在。通过功能相似性嵌入其候选产物，进一步证明轨迹仍停留在解空间的局部区域，我们称之为想法盆地。为了帮助智能体逃离这些盆地，我们研究了一种简单的周期性干预措施——fork-and-flush。我们的方法将智能体分叉为平行轨迹，每条轨迹继承累积的工作空间，但以全新的聊天上下文开始。在每条轨迹运行固定时间范围后，智能体从得分最高的轨迹继续。在13项长时程研究和工程任务中，单次智能体运行持续数天，fork-and-flush在同等计算预算下，相对于单次运行和最佳N基线，在最小-最大归一化平均分数上分别实现了66.0%和44.4%的相对提升。
英文摘要
Autoresearch agents tackle open-ended problems by repeatedly proposing candidate solutions, evaluating them, and using feedback to guide subsequent experiments. We show that independent runs of the same agent on the same task often plateau at substantially different scores, with gaps that persist even after considerable additional compute. Embedding their candidate artifacts by functional similarity provides further evidence that trajectories remain in localized regions of the solution space, which we call idea basins. To help agents escape these basins, we study a simple periodic intervention, fork-and-flush. Our method forks the agent into parallel trajectories, each inheriting the accumulated workspace but starting with a fresh chat context. After running each trajectory for a fixed horizon, the agent continues from the highest-scoring one. Across 13 long-horizon research and engineering tasks, with individual agent runs lasting up to several days, fork-and-flush outperformed the single-run and best-of-N baselines by a relative improvement of 66.0% and 44.4%, respectively, on the min-max normalized average score under an equal compute budget.
