Unverified50% confidenceFactExact time
When token representation similarity rises, the router's decision boundary becomes blurred, and multiple tokens are more easily mapped to the same set of experts, leading to decreased expert utilization and contraction of routing diversity.
1
Sources
50%
Confidence
Long-term
Relevance
10/16/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedIf the router consistently favors a small number of experts, the remaining experts will remain idle for extended periods, leading to low parameter utilization and unstable gradient updates.72% similarVerifiedError cascading refers to small judgment deviations in early steps being progressively amplified as the reasoning chain extends, ultimately producing results far from what was intended.72% similarUnverifiedThe performance gap for the same weights under different inference frameworks can sometimes reach more than ten percentage points.70% similarUnverifiedThe slow decline in divergence is traced to the smoothing of rare contexts, as increasing the amount of data cannot effectively reduce sampling variance in sparse regions.69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1041812API
curl https://kongchang.com/api/v1/knowledge/claims/1041812MCP
get_claim(id=1041812)