Unverified50% confidenceFactExact time
When performing k independent comparisons on the same dataset, the probability that at least one result by chance beats the baseline grows rapidly with k; this is known as the multiple comparisons problem or p-hacking.
1
Sources
50%
Confidence
Long-term
Relevance
10/12/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedExact K-nearest neighbor (KNN) requires traversing all candidate vectors, with time complexity growing linearly with data volume, which is often unacceptable at the scale of hundreds of millions.70% similarUnverifiedTop-p (nucleus sampling) limits the candidate token range by dynamically truncating probability mass, and has better adaptability than fixed-truncation Top-k70% similarUnverifiedP-hacking refers to the practice of researchers repeatedly adjusting data subsets, statistical methods, or inclusion criteria until the p-value falls below the significance threshold of 0.05, artificially creating the appearance of statistical significance.69% similarUnverifiedThe multiple testing problem refers to the phenomenon whereby the overall false positive rate rises sharply when multiple hypotheses are tested simultaneously; testing 20 capability dimensions at once means that on average 1 dimension will appear to show significant regression even if the agent has not changed.69% similarUnverifiedSelf-Consistency技术通过多次采样生成不同推理路径取最一致答案,代价是推理成本随采样次数线性增加68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1005189API
curl https://kongchang.com/api/v1/knowledge/claims/1005189MCP
get_claim(id=1005189)