Verified65% confidenceFactExact time
When the model decides whether to invoke a given Skill, it relies on semantic similarity matching and intent recognition, semantically comparing the request content against the descriptions of loaded Skills.
3
Sources
65%
Confidence
Long-term
Relevance
9/24/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedRelevance judgment is essentially a semantic similarity problem, where sentence vector models can be used to encode questions and actions, and then lightweight classifiers such as logistic regression or gradient boosting trees can be trained to output judgments.74% similarUnverifiedCognitive tasks typically require the model to output inferences about others' mental states, where the answers have relatively clear correct references and can be directly compared with human annotations; behavioral tasks require the model to choose among several candidate persuasion strategies, where the answer space is more structured and closer to pattern recognition and strategy retrieval.73% similarUnverifiedSkills 机制通过上下文注入在不修改模型权重的前提下引导模型输出，相比一次性 Prompt Engineering 强调可复用性、版本管理和模块化组合73% similarUnverifiedValue induction refers to the process of having a model acquire and express certain behavioral tendencies through specific linguistic samples during the post-training phase.72% similarUnverifiedBERT类模型在意图识别、情感分析、命名实体识别等判别式任务上具有优势，专注于理解而非生成72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/948047API
curl https://kongchang.com/api/v1/knowledge/claims/948047MCP
get_claim(id=948047)