引導式測驗 · 中等 等級
多代理編排測驗
在不假設可靠性的情況下測試多代理角色、委派、終止、通訊、框架範例和協調成本。
問題 1 的 8
多代理編排不是採用單一代理,而是設定什麼?
預覽每個問題8 問題
當您參加測驗時,就會出現答案和解釋。
- 多代理編排不是採用單一代理,而是設定什麼?
- 協調多個專業代理協作完成一項任務
- 無需工具即可運行一個非常大的模型
- 用單一微調模型取代所有代理
- 永久地將人類從每個決策中剔除
- In an orchestrator-worker design, what does the lead agent usually do?
- Runs all tools without a task plan
- Decomposes the task, delegates subtasks, and combines results
- Guarantees every worker is correct
- Replaces all termination checks
- Which is a real cost or risk of a multi-agent workflow?
- Guaranteed lower latency
- No need to check evidence
- More coordination overhead and opportunities for errors to propagate
- Automatic agreement with ground truth
- 終止條件在編排中扮演什麼角色?
- 它選擇模型的詞彙
- 它會對代理之間的訊息進行加密
- 它增加了使用的 GPU 數量
- 它防止代理在無限循環中交互
- Which multi-agent framework does the cited paper and documentation describe?
- 自動產生器
- LangGraph
- CrewAI
- 斯溫變壓器
- 集中式編排與分散式編排有何不同?
- 集中式使用GPU,分散式使用CPU
- 在集中控制中,一個協調器路由每一步;在去中心化中,代理直接交接
- 中心化沒有記憶,去中心化總是有記憶
- 它們在實踐中是相同的
- In AutoGen’s UserProxyAgent example, what may happen when an executable code block arrives and no human input is provided?
- The conversation is guaranteed to terminate successfully
- The agent silently rewrites its model weights
- The agent can execute the code and return the result as a message
- The assistant discards the code block and returns no message
- What communication model does the AutoGen paper use?
- A single static spreadsheet
- A one-way image classifier
- A shared GPU cache only
- Agents converse through customizable message-based interactions