让AI自动进化出更优的提示工程技能，提升大模型推理效果。
Meta Context Engineering via Agentic Skill Evolution
- 用双层智能体机制，让提示工程技能和上下文文件协同进化。
- 在5个不同领域测试中，相比顶尖方法平均提升16.9%，最高达53.8%。
- 适合需要高效、可迁移提示优化的AI研发与工程落地场景。
大型语言模型的推理效能高度依赖于其运行时上下文。这使得上下文工程（CE）成为优化输入的一门正式学科。当前的CE方法依赖人工设计的流程，如固定生成-反思工作流和预设上下文模板，存在结构偏见，将优化限制在狭窄且依赖直觉的设计空间内。为此，我们提出元上下文工程（MCE），一种双层框架，通过协同演化CE技能与上下文成果，取代静态的CE启发式方法。在MCE迭代中，元层智能体通过代理交叉（agentic crossover）——对技能历史、执行记录与评估结果进行深思熟虑的搜索——来优化工程技能；基础层智能体则执行这些技能，通过训练回放学习，并将上下文优化为灵活的文件与代码形式。我们在离线与在线设置下，于五个异构领域评估MCE。结果显示，相较于最先进的代理式CE方法，MCE实现了5.6%至53.8%的相对性能提升（均值16.9%），同时在上下文适应性、迁移能力及使用与训练效率方面表现更优。
原文摘要 · Abstract (English)
The operational efficacy of large language models relies heavily on their inference-time context. This has established Context Engineering (CE) as a formal discipline for optimizing these inputs. Current CE methods rely on manually crafted harnesses, such as rigid generation-reflection workflows and predefined context schemas. They impose structural biases and restrict context optimization to a narrow, intuition-bound design space. To address this, we introduce Meta Context Engineering (MCE), a bi-level framework that supersedes static CE heuristics by co-evolving CE skills and context artifacts. In MCE iterations, a meta-level agent refines engineering skills via agentic crossover, a deliberative search over the history of skills, their executions, and evaluations. A base-level agent executes these skills, learns from training rollouts, and optimizes context as flexible files and code. We evaluate MCE across five disparate domains under offline and online settings. MCE demonstrates consistent performance gains, achieving 5.6--53.8% relative improvement over state-of-the-art agentic CE methods (mean of 16.9%), while maintaining superior context adaptability, transferability, and efficiency in both context usage and training.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。