Unverified50% confidenceFactExact time
Weights tend to approximate a symmetric distribution and are suitable for symmetric quantization, while activations, after passing through nonlinear functions such as ReLU, exhibit shifted distributions more suitable for asymmetric quantization.
1
Sources
50%
Confidence
Long-term
Relevance
8/7/2026
First Seen
Sources
图解量化：大语言模型压缩核心技术详解
Related Claims
Unverified旋转变换量化的理论基础是正交矩阵不改变模型数学等价性但能重新分布数值统计特性，用于搅匀离群值68% similarUnverifiedThe Bernoulli distribution belongs to the exponential family, and the mapping between its natural parameter and mean parameter is exactly the sigmoid function; logistic regression is the canonical link function choice for the Bernoulli distribution within the GLM framework.68% similarUnverifiedRL对pass@1的提升而不影响pass@k，本质上是在压缩模型的输出分布使其更集中在正确答案附近，即分布锐化61% similarUnverified在实时增强逻辑下，关键不在于固定的组合数量，而在于增强的随机分布范围与参数强度61% similarUnverified贝叶斯网络通过有向无环图建模变量间条件依赖关系，Copula函数将多个边缘分布耦合为联合分布，两者对高维数据的扩展性较差61% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/705552API
curl https://kongchang.com/api/v1/knowledge/claims/705552MCP
get_claim(id=705552)