Unverified50% confidenceOpinionExact time
Decorative confidence numbers tend to be high and narrow, barely changing with question difficulty, whereas measurement-based numbers checked against outcomes are allowed to drop on genuinely close questions.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/15/2026
First Seen
Valid until: 1/13/2027
Sources
Related Entities
Related Claims
UnverifiedThe harmonic mean is sensitive to weak links; as long as performance in any one dimension is poor, the overall score will be noticeably pulled down.74% similarUnverifiedMean rather than median should be used to estimate on-device throughput, because the median filters out slow but real use cases, making capacity estimates overly optimistic.73% similarUnverifiedQuantization typically brings a 1%–3% accuracy loss, but for discriminative tasks with a limited label space such as intent classification it is almost imperceptible.71% similarUnverifiedThe standard error scales with the square root of n, so artificially increasing n narrows the error bars, leading to spurious statistical significance.71% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1037017API
curl https://kongchang.com/api/v1/knowledge/claims/1037017MCP
get_claim(id=1037017)