Unverified50% confidenceConceptExact time
Gemma 4 E2B's embedding table occupies the vast majority of the model weights, because the vocabulary usually has tens of thousands to hundreds of thousands of entries, each with a dimensionality of several thousand, causing the embedding table to easily occupy several GiB, so the quantization strategy must prioritize the embedding.
1
Sources
50%
Confidence
Long-term
Relevance
10/12/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedImplicit solving requires solving a system of linear equations, with each step requiring second-order information similar to the Hessian or its approximation, resulting in enormous computational and storage overhead at the scale of millions or billions of parameters.58% similarVerifiedElastic Weight Consolidation (EWC) estimates the importance of parameters to old tasks via the Fisher information matrix and applies stronger regularization penalties to important parameters to mitigate catastrophic forgetting.58% similarUnverifiedEmbeddingGemma 2 is a variant fine-tuned on top of Gemma 2 through contrastive learning, with the goal of making the vector distance between a query and the correct option as small as possible and the distance to incorrect options as large as possible.57% similarUnverified当前主流模型拥有数万至数十万token的上下文，但长篇连载创作中角色设定等细节仍面临漂移风险56% similarUnverifiedVRAM capacity often determines what size model can be loaded more than raw compute performance does.56% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1011107API
curl https://kongchang.com/api/v1/knowledge/claims/1011107MCP
get_claim(id=1011107)