Unverified50% confidenceTradeoffExact time
A hierarchical path prefix dilutes the weight of the body text and slightly lengthens the sequence, so the path should not be too deep; generally the most recent two or three levels are kept
1
Sources
50%
Confidence
Long-term
Relevance
10/13/2026
First Seen
Sources
Related Entities
Related Claims
Unverified路由层不压缩的工程依据是：路由错误会在整个前向传播中级联放大且无法后续纠正，而路由层参数量占比极低69% similarUnverifiedIn multi-hop scenarios, intermediate nodes are semantically distant from the original question and are not adjacent in the vector space, causing the retriever to fail to include them in the candidate set.66% similarUnverifiedRouting an LLM by total prompt length (directing requests below approximately 500 characters to the economy tier and above 500 to the premium tier) causes short prompts carrying tool definitions to be incorrectly routed to lower-capability models.65% similarUnverifiedA fallback chain should be kept short — three models is generally the upper limit — as the worst-case path latency worsens beyond that.65% similarUnverifiedThe limitation of depth scaling is that if the initial reasoning direction is wrong, a deeper reasoning chain will only go further and further down the wrong path.65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1019563API
curl https://kongchang.com/api/v1/knowledge/claims/1019563MCP
get_claim(id=1019563)