Unverified50% confidenceFactExact time
Anthropic 提出 Constitutional AI 框架约束模型行为，OpenAI 在 Agent 产品中引入 human-in-the-loop 机制，Google DeepMind 发表了关于 Agent 对齐的研究论文
1
Sources
50%
Confidence
Long-term
Relevance
8/18/2026
First Seen
Sources
Related Claims
UnverifiedOpenAI's Function Calling and Anthropic Claude's Tool Use are representative implementations of the Agent pattern.73% similarUnverifiedThe classic definition of Agent can be traced back to Russell and Norvig's textbook "Artificial Intelligence: A Modern Approach"73% similarUnverifiedRepresentative Agent tools include Google AI Studio, Manus, and Coze.72% similarUnverifiedMainstream AI Agent frameworks adopt the Human-in-the-Loop principle, mandating that human confirmation be awaited at critical decision nodes.72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/767538API
curl https://kongchang.com/api/v1/knowledge/claims/767538MCP
get_claim(id=767538)