Auto-sync: 2026-04-17 16:17

This commit is contained in:
2026-04-17 16:17:32 +08:00
parent 2935282989
commit a1c4f6495b
17 changed files with 525 additions and 17 deletions

View File

@@ -0,0 +1,29 @@
---
title: "Preference Learning"
type: concept
tags: [ai, personalization, machine-learning]
last_updated: 2026-04-17
---
## Definition
AI Agent 通过与用户交互逐渐学习并记住用户偏好,用于持续优化内容筛选和推荐结果。
## How It Works
1. 初始交互:用户设定偏好规则(如"不包含表情包""喜欢技术类内容"
2. 反馈收集:每次呈现结果后询问用户是否满意
3. 规则更新:根据用户反馈调整偏好规则
4. 持续优化:随着时间推移,筛选结果越来越符合用户需求
## Use Cases
- [[Daily Reddit Digest]]:学习用户感兴趣的子版块和内容类型
- [[Custom Morning Brief]]:根据用户阅读习惯调整新闻优先级
- 内容推荐系统:个性化推荐引擎的核心机制
## Connections
- [[Task Automation]] ← enables ← [[Preference Learning]]
- [[Context Memory]] ← stores ← [[Preference Learning]]
- [[Cron Jobs]] ← schedules ← [[Preference Learning]]
## References
- 通过 Memory 目录存储用户偏好规则
- 每次交互后更新规则,形成反馈循环