Unverified50% confidenceFactExact time
For unimodal distributions, the mode is insensitive to skewness and heavy tails of the distribution, because the peak location is mainly determined by local density, and outliers do not pull the mode the way they pull the mean.
1
Sources
50%
Confidence
Long-term
Relevance
10/27/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe compression ratio is calculated as (original context − compressed context) / original context; it is a ratio-type metric that is extremely sensitive to extreme samples, so the average is easily dominated by a few extreme values, and the median and quantile distribution should be examined.60% similarUnverified平均值的部分聚合必须同时保存总和与计数，因为不同分区的平均值不能简单再求平均，必须按样本数加权58% similarUnverifiedBranch pruning performs backpropagation only on the sampled preference branch, restricting gradient computation to a single branch.57% similarUnverifiedVendors no longer simply pursue the expansion of parameter scale, but instead pay more attention to the intelligence density per unit token.57% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1102401API
curl https://kongchang.com/api/v1/knowledge/claims/1102401MCP
get_claim(id=1102401)