Unverified50% confidenceSolutionExact time
在提示中提供对比性示例（同时给出正反案例）比单独给出正面案例的引导效果更强，属于少样本学习（Few-shot Learning）
1
Sources
50%
Confidence
Long-term
Relevance
8/25/2026
First Seen
Sources
Related Entities
Related Claims
Unverified样例学习效应表明对初学者而言学习已解决的例题比自主解题更有效79% similarUnverified增加负样本描述（When NOT to Use）能明确 Skill 的决策边界，显著减少误触发71% similarUnverified在面试作业中应用实验数据证明简单方案优于复杂方案，而非仅凭经验断言70% similarUnverifiedThe root cause of rising hallucination rates in reasoning models is misaligned incentives: reinforcement learning training causes models to tend to generate detailed and persuasive chains of thought, and human evaluators give higher scores to more elaborately argued answers, even when they contain factual errors.69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/797942API
curl https://kongchang.com/api/v1/knowledge/claims/797942MCP
get_claim(id=797942)