Unverified50% confidenceFactExact time
Classic skill discovery methods such as the Options Framework decompose policies into two levels: high-level option selection and low-level option execution; DIAYN and VALOR unsupervisedly emerge skills by maximizing information-theoretic objectives for skill diversity.
1
Sources
50%
Confidence
Long-term
Relevance
9/11/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedSkill推荐架构中通过余弦相似度检索最相关Skill候选集，再由LLM进行重排序并决策调用顺序75% similarVerifiedSkills采用渐进式披露设计，平时只向模型提供技能的名称和描述，具体执行规则在需要时才加载73% similarUnverifiedSkills范式通过系统提示词和结构化指令将专业方法论嵌入Agent决策框架，优势在于门槛低、可版本控制、易于社区协作，无需额外模型训练成本72% similarUnverifiedSkills机制与LangChain的Tool Calling、OpenAI的Function Calling同属一脉，核心思想是将LLM推理能力与特定领域工具解耦70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/898863API
curl https://kongchang.com/api/v1/knowledge/claims/898863MCP
get_claim(id=898863)