新论文提出护栏式持续学习，防止智能体遗忘
Banger paper on harness continual learning.
Banger paper on harness continual learning.
关于驾驭持续学习的重磅论文。
(bookmark it)
（收藏它）
If you already are allowing your agents to rewrite their own prompts, skills, or memory files, this one is worth your time.
如果你已经允许你的代理重写自己的提示、技能或记忆文件，这篇值得你花时间阅读。
(bookmark it)
（收藏它）
Continual learning has always tracked what changes in the weights. Modern agents accumulate experience in the harness instead, across prompts, memories, tools, skills, and routing rules.
持续学习一直追踪权重中的变化。现代代理则相反，在驾驭层中积累经验，跨越提示、记忆、工具、技能和路由规则。
What this means is that if you update any harness component, previously reliable behavior can break with the model completely untouched. The paper names that harness-level forgetting and provides a way to measure it.
这意味着，如果你更新任何驾驭组件，即使模型完全未动，之前可靠的行为也可能被破坏。论文将此命名为驾驭层遗忘，并提供了一种测量方法。
Guarded harness evolution separates proposing an update from committing it. A Continual Optimizer drafts a candidate harness from post-execution feedback, and a Continual Evaluator commits only after checking current improvement, historical retention, and validity.
有保护的驾驭进化将提出更新与提交更新分离。持续优化器根据执行后反馈起草候选驾驭，持续评估器仅在检查当前改进、历史保留和有效性后才提交。
Relative gains exceed 10% across textual reasoning, multimodal perception, and open-world interaction.
在文本推理、多模态感知和开放世界交互中，相对增益超过10%。
Paper: https://arxiv.org/abs/2608.19013
论文：https://arxiv.org/abs/2608.19013
Track more trending AI papers in our academy: https://academy.dair.ai/
在我们的学院中追踪更多热门AI论文：https://academy.dair.ai/
更进一步：量化金融体系
看懂新闻只是起点——沿量化金融路径，把它变成能交付的工程能力