Unverified50% confidenceFactExact time
大模型默认输出的是自然语言文本，而实际业务中往往需要结构化数据（如JSON）以便程序处理
1
Sources
50%
Confidence
Long-term
Relevance
9/11/2026
First Seen
Sources
Related Entities
Related Claims
Unverified结构化JSON格式的上下文往往比自然语言描述带来更稳定的模型输出79% similarUnverified大语言模型的自然倾向是生成流畅文本，而Agent工作流需要严格、可预测、机器可读的结构化输出，两者存在冲突76% similarUnverified大语言模型在处理复杂或冗长的结构时容易出现类型退化，会把本应是数组的字段序列化成JSON字符串发送76% similarUnverified大语言模型可通过Zero-shot方式同时完成意图识别与槽位提取，并直接输出为JSON等结构化格式73% similarUnverifiedLarge language models (LLMs) trained on massive code repositories and natural language corpora can map human natural language intent into executable code, a capability known as Natural Language Programming.73% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/896515API
curl https://kongchang.com/api/v1/knowledge/claims/896515MCP
get_claim(id=896515)