Google DeepMind 论文：自我进化智能体边学边忘，学新任务最快的老任务掉分最多
DeepMind 实测了三种让智能体自我进化的方法，发现学新任务最快的老任务掉分最多，GEPA 一个 15.5 分的收益被改写后只剩 2.5 分，做智能体的都该看看。
Google DeepMind 发布论文，研究通过编辑系统提示词、记忆或技能文件实现自我进化的智能体是否会在学习新任务时损害旧任务。团队在主题、工作流、领域不断变化的任务流上测试了 ReasoningBank、GEPA 和 Meta-Harness 三种方法。GEPA 平均每个区块新学 6.2 分，同时在其他任务上平均丢失 4.3 分；它学会向用户询问缺失细节后获得 15.5 分提升，但后续区块改写该规则后收益跌至 2.5 分。论文建议将任务专属规则分开存放，并在每次更新后重测旧任务。
New Google DeepMind paper finds that self-improving agents forget as they learn, and the method that learned new tasks fastest broke old ones the most.
Self-evolving agents improve without retraining by editing their system prompt, memory, or skill files. In real use tasks keep shifting, and each edit made for 1 task also applies to every task that follows.
The team ran ReasoningBank, GEPA, and Meta-Harness on task streams that changed topic, workflow, or domain. They tracked how much each method learned on the current task and lost on the others.
GEPA learned the most, 6.2 points per block on average, and also lost the most elsewhere, 4.3 points. After it learned to ask users for missing details, a later block rewrote that rule and its 15.5-point gain fell to 2.5.
If your agent edits its own prompt or memory, keep task-specific rules separate and retest old tasks after every update.
- arXiv: DeepSeek10-19 08:45原文
