Unverified50% confidenceFactExact time
The joint training of this method uses only cross-entropy loss, simultaneously accounting for both accuracy and calibration quality, at the cost of introducing only one additional hyperparameter: the calibrator learning rate.
1
Sources
50%
Confidence
Long-term
Relevance
10/12/2026
First Seen
Sources
Related Entities
Related Claims
Unverified混合精度训练在不显著损失模型精度的前提下，可将训练速度提升约2倍，显存占用降低近一半75% similarUnverifiedMaintaining higher entropy in early training and gradually annealing the entropy coefficient downward as training progresses helps agents escape local optima.73% similarUnverifiedScaling Law主要适用于预训练损失（交叉熵）这一代理指标，而非所有下游任务的实际性能72% similarUnverified计算最优（Compute-Optimal）并不等于集群最优（Cluster-Optimal），真正决定训练效率的因素远不止浮点运算次数72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1002292API
curl https://kongchang.com/api/v1/knowledge/claims/1002292MCP
get_claim(id=1002292)