Unverified50% confidenceOpinionExact time
将薄弱或不完整的文本块交给小模型，输出质量会迅速崩坏
1
Sources
50%
Confidence
Long-term
Relevance
8/27/2026
First Seen
Sources
Related Entities
Related Claims
Unverified实验表明，仅经过数代自我训练，模型输出的方差就会急剧缩小，文本趋于同质化73% similarUnverified块太大会导致嵌入向量稀释、检索精度下降，块太小则可能丢失完整的语义单元71% similarUnverified块太小会导致缺乏上下文，块太大会引入噪声并受限于模型上下文窗口长度，切块参数存在权衡69% similarUnverified小参数模型在代码生成场景下的本质短板是精度不足，表现为函数未声明、空函数体、工具调用格式损坏、变量拼写错误等69% similarUnverified文档分块策略直接影响检索召回率，块太大会引入噪声，块太小则丢失语义完整性68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/810050API
curl https://kongchang.com/api/v1/knowledge/claims/810050MCP
get_claim(id=810050)