Unverified50% confidenceDecision RuleExact time
The model that writes the answer should not decide whether to answer; every message first goes through a separate cheap classification call that returns one of three labels: in-scope answer, out-of-scope refusal, or handoff to a human.
1
Sources
50%
Confidence
Long-term
Relevance
10/15/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedRefusal and handoff replies use fixed strings rather than model generation, to reduce the attack surface for prompt injection.69% similarUnverifiedWhen few-shot examples demonstrating label contrasts are included in the prompt, all large language models outperform zero-shot prompting.66% similarUnverified受限解码在保证格式正确性的同时，可能改变模型实际采样分布，导致语义质量的次优选择65% similarUnverified在提示词中要求模型简短无法降低输入成本，因为上下文在读取提示词之前就已发送64% similarUnverifiedLarge language models respond systematically better to concrete few-shot examples than to abstract style labels, and providing actual email copy is more effective at locking in output style than abstract descriptions.64% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1034981API
curl https://kongchang.com/api/v1/knowledge/claims/1034981MCP
get_claim(id=1034981)