Unverified50% confidenceDecision RuleExact time
分块大小通常推荐在 256-512 tokens 之间,并保留 10-15% 的重叠区域
1
Sources
50%
Confidence
Long-term
Relevance
9/11/2026
First Seen
Sources
Related Entities
Related Claims
VerifiedRAG实践中chunk_size通常设置在256至1024 token之间,并保留约10%至20%的重叠(Overlap)77% similarUnverified实际工程中Chunk大小通常为200-1000 tokens,Overlap比例通常为10%-20%,是需要根据文档类型和业务场景调优的关键超参数77% similarUnverified实际工程中chunk_size在256至1024 token之间、chunk_overlap在20至100 token之间是较为常见的经验范围76% similarUnverified作者认为12-30 token/s是舒适区,超过30 token/s算好;48-64GB的M5 Max约能跑到20 token/s76% similarUnverified知识库 Chunk Size 分块大小的业界通行经验值通常在 512 至 1024 Token 之间,并辅以约 20% 的重叠窗口以保持语义连贯性70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/905009API
curl https://kongchang.com/api/v1/knowledge/claims/905009MCP
get_claim(id=905009)