能力驱动的智能体记忆自进化

中文摘要 AI 辅助

记忆自进化利用任务反馈迭代改进可执行的记忆程序，这些程序负责存储和检索过去交互中的信息。现有方法通常采用整体进化，从混合反馈中推导修订方向，并以整体性能判断进展。这可能会模糊优化方向，并隐藏因其他方面性能回退而被抵消的特定能力提升，导致有前景的方向未被充分探索。我们引入了能力驱动的进化，将搜索指导从整体性能扩展到个体能力维度，保留有前景的修订，并将探索范围扩展到整体进化的边界之外。我们提出了PrisMem，它使用依赖感知的能力选择来优先考虑具有潜在跨能力收益的目标，并通过历史引导的诊断来优化能力专家。迹引导的整合在成对的差异案例上比较评估过的程序，利用它们的行为差异将互补的收益整合到一个统一的记忆程序中。实验表明，PrisMem在BEAM-1M和LongMemEval-M上分别比最强基线高出10.54和7.83个百分点，证明了其在百万级token历史中的有效性。

英文摘要

Memory self-evolution uses task feedback to iteratively improve executable memory programs that store and retrieve information from past interactions. Existing approaches typically adopt holistic evolution, deriving revision directions from mixed feedback and judging progress by overall performance. This can obscure optimization directions and hide capability-specific gains offset by regressions elsewhere, leaving promising directions underexplored. We introduce capability-driven evolution, which extends search guidance from overall performance to individual capability dimensions, preserving promising revisions and expanding exploration beyond the boundaries of holistic evolution. We propose PrisMem, which uses dependency-aware capability selection to prioritize targets with potential cross-capability benefits and history-guided diagnosis to refine capability specialists. Trace-guided integration compares evaluated programs on paired differential cases, using their behavioral differences to consolidate complementary gains into a unified memory program. Experiments show that PrisMem outperforms the strongest baselines by 10.54 and 7.83 percentage points on BEAM-1M and LongMemEval-M, respectively, demonstrating its effectiveness on million-token histories.

发表机构

• University of Science and Technology of China（中国科学技术大学）
• Microsoft（微软）
• University of California, San Diego（加利福尼亚大学圣迭戈分校）

机构由 AI 辅助整理，请以论文原文为准。