Unverified50% confidenceFactExact time
Harvey, Liu & Zhu（2016）在《Journal of Finance》的研究表明大量已发表的因子研究存在数据挖掘偏差，有效t统计量门槛应提高到3.0以上
1
Sources
50%
Confidence
Long-term
Relevance
7/21/2026
First Seen
Sources
Related Claims
Unverified学术论文发表的因子在发布后平均衰减约32%（Harvey & Liu, 2020）61% similarUnverifiedDeepMind于2022年的Chinchilla论文修正了Scaling比例，发现相同计算预算下应使用更小模型但喂入更多数据61% similarUnverified研究表明即使是轻微的数据污染也可能导致评测分数虚高5-15个百分点60% similarUnverifiedMultiple studies in 2023 found that abnormally high scores from certain models on specific benchmarks were highly correlated with data leakage60% similarUnverified2022年Hoffman等人的Chinchilla论文重新校准了算力最优的参数-数据比例，指出此前许多大模型训练不充分59% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/577437API
curl https://kongchang.com/api/v1/knowledge/claims/577437MCP
get_claim(id=577437)