Unverified50% confidenceConceptExact time
μP (Maximal Update Parametrization) is a scaling theory about how neural networks are parameterized, whose core goal is to keep optimal hyperparameters (especially the learning rate) stable across different model widths, thereby enabling zero-shot transfer.
1
Sources
50%
Confidence
Long-term
Relevance
10/19/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedGreg Yang等人提出的μP（Maximal Update Parameterization）理论试图在保留特征学习的同时实现理论分析71% similarVerifiedModel quantization is a technique that compresses neural network weights from high-precision floating-point numbers to low-precision integers.65% similarVerifiedSpeculative decoding is a lossless acceleration technique that does not alter the model's output distribution; it uses a draft model with far fewer parameters than the target model to generate candidate tokens, which the target model then verifies in parallel.65% similarUnverified模型重参数化是指在不改变网络输出函数的前提下，通过数学等价变换调整计算图结构的技术，严格保证变换前后输出完全一致，不需要任何微调。Priming借用了重参数化的思想，但不要求数学等价，因此需要少量微调。65% similarUnverifiedFull fine-tuning updates all weights of the model, yielding the best results but requiring extremely high computing power63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1068345API
curl https://kongchang.com/api/v1/knowledge/claims/1068345MCP
get_claim(id=1068345)