Unverified50% confidenceBenchmarkExact time
The study fitted a power-law scaling relationship for the generalization onset time as T_grok ∝ H^(-0.27) D^(-2.04) η^(-0.50) λ^(-0.64), where H is model width, D is data complexity, η is learning rate, and λ is weight decay.
1
Sources
50%
Confidence
Long-term
Relevance
9/12/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedCohen's d的计算公式为 d = (μ₁ - μ₂) / σ_pooled，其中σ_pooled是两组样本的合并标准差64% similarUnverified卡尔曼增益公式为 K = PH^T(HPH^T + R)^{-1}，在线性高斯条件下达到理论最优估计63% similarUnverifiedKaplan et al. found that the relationship between model loss and parameter count N approximately satisfies the power-law relationship L(N) ∝ N^(-0.076).61% similarUnverifiedThe mathematical expression for Ledoit-Wolf shrinkage is Σ_shrunk = δ·F + (1-δ)·S, where S is the sample covariance matrix, F is the shrinkage target, and δ∈[0,1] is the shrinkage intensity.60% similarUnverifiedThe core issue involved in the discovery is a factorial normalization discrepancy between the original moment M_n and the Taylor/Jensen coefficients.59% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/916330API
curl https://kongchang.com/api/v1/knowledge/claims/916330MCP
get_claim(id=916330)