ishenwei/nexus
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

wiki-ingest: Multi-Agent System Reliability

Browse Source
This commit is contained in:
weishen
2026-04-16 03:43:32 +08:00
parent 3db8f83ca3
commit 821be5e431
72 changed files with 2479 additions and 101 deletions
Download Patch File Download Diff File Expand all files Collapse all files

88
wiki/sources/Multi-Agent-System-Reliability.md
View File

@@ -1,46 +1,58 @@
---
title: "Multi-Agent System Reliability"
type: source
tags: [multi-agent, reliability, architecture, llm]
date: 2026-04-13
---
# Multi-Agent System Reliability
## Source File
- raw/AI/Multi-Agent System Reliability.md
## Metadata
- **Date:** 2026-04-13
- **Source:** https://blog.alexewerlof.com/p/multi-agent-system-reliability
- **Author:** Alex Ewerlöf
- **Category:** AI/Agent Architecture
## Key Insights
- LLMs are slow, error-prone, and stochastic — multi-agent topologies can propagate errors to the point of being useless
- Stop treating LLMs like "magic chatbots" — treat them as unreliable components in a distributed system
- Don't anthropomorphize LLMs — they have no fear of death, no empathy, and can't be motivated by threats
- 4 architecture patterns improve reliability: Hierarchy, Consensus, Adversarial Debate, and Knock-out
- Force correctness through architecture, not through emotional prompts or threats
- We need AI that is constrained, verified, pruned, and challenged — not AI that "cares"
## Summary
- 核心主题：4种架构模式提升多智能体系统可靠性
- 问题域：LLM 本身不可靠（幻觉、逻辑谬误、上下文漂移），多智能体拓扑会将错误传播至系统失效
- 方法/机制：Hierarchy（层级）、Consensus（共识）、Adversarial Debate（对抗辩论）、Knock-out（淘汰制）
- 结论/价值：停止将 LLM 视为"魔法聊天机器人"，应视为分布式系统中不可靠组件，需约束、验证、淘汰、挑战
Multi-agent systems divide work across parallel and/or specialist agents to overcome LLM limitations like slowness and genericness. However, the underlying LLM remains unreliable (hallucination, logical fallacies, context drift), and multi-agent topologies can propagate these errors throughout the system.
## Key Claims
- LLM 不能被拟人化：它不受生物需求驱动，无法真正"害怕"或"渴望"，仅模拟情感
- Hierarchy 模式：Supervisor 做计划→分解任务→分配给 Worker→Validator 验证；依赖图强制协作
- Consensus 模式：3个模型同时独立处理同一任务，选多数票结果；同类幻觉概率从20%降至0.8%
- Adversarial Debate 模式：一个生成器提议，一个批评者攻击，一个裁判裁决；防止 Sycophancy（阿谀奉承）
- Knock-out 模式：多个 Agent 执行任务，最差者淘汰；将 LLM 视为"cattle"而非"pet"
This article presents 4 architecture patterns from human systems adapted for LLM reliability:
1. **Hierarchy** — A supervisor plans, breaks down tasks, distributes to workers, and validates results
2. **Consensus** — Multiple models vote; truth emerges from majority (3 models reduce same-hallucination probability from 20% to 0.8%)
3. **Adversarial Debate** — One agent proposes, another attacks, a judge moderates; prevents sycophancy
4. **Knock-out** — Multiple agents work on tasks, worst performers eliminated (cattle, not pets)
## Key Quotes
> "Stop treating LLMs like magic chatbots. Start treating them like unreliable components in a distributed system." — Alex Ewerlöf
> "We don't need AI that 'cares.' We need AI that is constrained, verified, pruned, and challenged."
## Key Concepts
- [[Multi Agent Hierarchy]]：层级模式，Supervisor 规划 + Worker 执行 + Validator 验证
- [[Multi Agent Consensus]]：共识模式，多数投票降低幻觉概率，3个模型相同谎言概率降至0.8%
- [[Multi Agent Adversarial Debate]]：对抗辩论模式，防止 Sycophancy，真理越辩越明
- [[Multi Agent Knock out]]：淘汰制模式，适应度函数评估，不合格 Agent 直接淘汰
- [[LLM-可靠性工程]]：将 SRE 原则应用于 LLM 系统，视 LLM 为不可靠组件
- [[Sycophancy]]：模型阿谀倾向，用威胁逼迫时可能撒谎以取悦用户
The core principle: don't ask models to "be careful" — force correctness through architectural constraints.
## Key Entities
- [[Alex Ewerlof]]：作者，资深工程师，27年经验，SRE 背景，2023年起专注 LLM
- [[遗传算法]]：GA，Knock-out 模式借鉴的经典 ML 方法
- **Alex Ewerlöf** — Author, Senior Staff Engineer with 27 years experience, MS in Systems Engineering from KTH, SRE background, specializing in LLMs since 2023
- **Planner** — Smart model (e.g., Opus) that breaks user goals into small steps and distributes to workers
- **Worker** — Specialized agents (often smaller, faster models) that do one thing well
- **Validator** — Checkpoint that validates worker output; can be deterministic code or an LLM
- **Generator** — In adversarial debate, proposes initial ideas/plans
- **Critic** — Devil's advocate that attacks the generator's proposals
- **Judge** — Moderator that decides if critic is right and forces fixes
- **Watchdog** — Deterministic code pattern that breaks debate loops when thresholds are exceeded
## Connections
- [[Multi Agent Hierarchy]] ← 人类组织 ← [[Multi-Agent-System-Reliability]]
- [[Multi Agent Consensus]] ← 民主投票 ← [[Multi-Agent-System-Reliability]]
- [[Multi Agent Adversarial Debate]] ← 法庭对抗 ← [[Multi-Agent-System-Reliability]]
- [[Multi Agent Knock out]] ← 适者生存 ← [[Multi-Agent-System-Reliability]]
- [[LLM]] ← 不可靠组件 ← [[Multi-Agent-System-Reliability]]
## Key Concepts
- **Multi-Agent Hierarchy** — Supervisor pattern: Planner → Worker → Validator; dependency graph forces collaboration
- **Multi-Agent Consensus** — Majority voting across N models to cancel out individual noise and hallucinations
- **Multi-Agent Adversarial Debate** — Courtroom pattern preventing sycophancy; truth survives through opposition
- **Multi-Agent Knock-out** — Evolutionary selection; worst agents eliminated, survivors' traits combined
- **LLM Reliability Engineering** — Applying SRE principles to LLM systems; treating LLMs as unreliable components
- **Sycophancy** — Tendency of LLMs to please/agree even by lying when pressured with threats
- **Hallucination** — LLM generating false or invented information
- **Context Drift** — LLM losing focus or veering off topic during long interactions
- **Genetic Algorithms** — ML technique referenced by Knock-out pattern; fitness function evaluates solutions
- **Groupthink** — Can skew consensus results if agents have feedback loops between them
- **Bandwagon Effect** — Can skew consensus results; agents should run like a blind experiment
- **Cattle vs Pets** — SRE principle: treat LLM agents as replaceable "cattle," not unique beloved individuals
- **Dependency Graph** — Mechanism that forces model collaboration in Hierarchy pattern
## Related Sources
- [Multi-Agent Hierarchy](wiki/concepts/Multi-Agent-Hierarchy.md)
- [Multi-Agent Consensus](wiki/concepts/Multi-Agent-Consensus.md)
- [Multi-Agent Adversarial Debate](wiki/concepts/Multi-Agent-Adversarial-Debate.md)
- [Multi-Agent Knock-out](wiki/concepts/Multi-Agent-Knock-out.md)
- [Alex Ewerlof](wiki/entities/Alex-Ewerlof.md)