Unverified50% confidenceFactExact time
A 'replication crisis' exists in the machine learning field, with a considerable proportion of paper results being impossible to fully reproduce.
1
Sources
50%
Confidence
Long-term
Relevance
9/1/2026
First Seen
Sources
Related Entities
Related Claims
Unverified科研领域的复现危机根源之一是分析代码与原始数据的不透明68% similarUnverified2022年的一项系统性研究发现，大量 PINN 相关论文难以从论文描述独立复现，部分原因在于该领域尚未形成成熟的代码发布规范63% similarUnverifiedMarcos López de Prado noted in Advances in Financial Machine Learning that most trading strategies published in academic journals are not statistically significant after controlling for multiple testing.60% similarUnverifiedLarge language models and machine learning models are built on statistical patterns in historical data and often fail in the face of structural break events such as the COVID-19 liquidity crisis in March 2020 and the UK pension crisis in 2022.59% similarUnverified金融数据的非平稳性意味着消融结论在一个时段成立未必在另一时段同样有效，这是金融AI消融实验的独特挑战59% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/838962API
curl https://kongchang.com/api/v1/knowledge/claims/838962MCP
get_claim(id=838962)