Unverified50% confidenceBenchmarkExact time
Under exactly the same initialization, data and learning rate schedule, the research team trained the same model from scratch on a 32-chip NPU and on a GPU respectively, and the Pearson correlation coefficient between the two loss curves reached 0.99.
1
Sources
50%
Confidence
Long-term
Relevance
10/14/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe trained model achieves a per-token negative log-likelihood (NLL) of 0.93, compared to 12.60 for a randomly initialized twin model of the same architecture.71% similarVerifiedIn a reproduction experiment training for 40 epochs on CIFAR-10, plain-56 achieved a training accuracy of 84.0% and plain-20 achieved a training accuracy of 95.1%, even though plain-56's parameter count (853,018) is approximately three times that of plain-20 (269,722).66% similarUnverifiedThe power-law exponent between test loss and compute is approximately 0.05, meaning that every 10x increase in compute yields approximately a 12% improvement in model performance.65% similarUnverifiedAt a training scale approaching 100 billion tokens, the difference in final smoothed training loss between the polynomial version and the native implementation falls between -0.107 and +0.079.64% similarUnverifiedAdam 优化器由 Kingma 和 Ba 于 2014 年提出，同时维护梯度的一阶矩估计与二阶矩估计，并通过偏差修正项处理训练初期矩估计偏向零的问题64% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1026570API
curl https://kongchang.com/api/v1/knowledge/claims/1026570MCP
get_claim(id=1026570)