Files
nexus/wiki/concepts/Multi-Agent-Knock-out.md
2026-04-15 15:02:52 +08:00

38 lines
1.3 KiB
Markdown
Raw Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
---
title: "Multi-Agent Knock-out"
type: concept
tags: [multi-agent, architecture, reliability, genetic-algorithm]
last_updated: 2026-04-15
---
## Definition
一种多智能体架构模式借鉴遗传算法GA多个 Agent 执行任务,适应度函数评估,最差者被淘汰。核心思想:用"适者生存"替代"死亡恐惧"。
## How It Works
1. 将任务分配给 N 个 Agent
2. Validator适应度函数决定哪些 Agent 被淘汰
3. (可选)用获胜 Agent 的特征组合生成新 Agent 填补空缺
## SRE 类比
- LLM Agent = "cattle"(牲畜,可替换)
- 不给它命名期待它做好:启动 → 检查 → 失败则淘汰
## Key Requirements
- 必须有快速验证输出的方式(如单元测试)
- 若需人工检查所有分支则太慢,此模式优势消失
## Genetic Algorithm Connection
借鉴传统 ML 的遗传算法两个要素:
- **遗传表示**:模型及其上下文
- **适应度函数**:淘汰函数
## Best For
- 迭代式 Agent 工程开发
- 调试阶段,不适合生产环境和大用户负载
## Related Concepts
- [[Multi-Agent-Hierarchy]]:层级验证模式
- [[Multi-Agent-Consensus]]:投票共识模式
- [[Multi-Agent-Adversarial-Debate]]:对抗辩论模式
- [[遗传算法]]:本模式借鉴的经典 ML 方法