Unverified50% confidenceFactExact time
The concept of Scaling Laws was first systematically proposed by institutions such as OpenAI and DeepMind in research around 2020, with the core finding that model cross-entropy loss exhibits a smooth power-law relationship with parameter count, training data volume, and compute, and shows no obvious signs of saturation over a fairly wide range.
1
Sources
50%
Confidence
Long-term
Relevance
10/16/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedScaling Laws表明模型的交叉熵损失会随着计算量、数据量和参数量的增加而平滑下降，且这种下降关系可以用简单的幂律函数精确拟合76% similarUnverifiedOpenAI、DeepMind等机构的研究表明模型性能随参数量、数据量和计算量的增加呈幂律提升（Scaling Law）70% similarUnverifiedScaling Laws describe the decrease in pre-training loss, not a linear improvement in all downstream capabilities; some capabilities, such as complex reasoning, exhibit emergent properties.69% similarUnverifiedThe paper derives an excess risk bound to characterize the generalization performance of the proposed method, which is a theoretical tool that end-to-end deep models typically cannot provide.67% similarUnverified双重下降现象为OpenAI等机构提出的神经网络缩放定律（Scaling Laws）提供了理论基础66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1047934API
curl https://kongchang.com/api/v1/knowledge/claims/1047934MCP
get_claim(id=1047934)