Unverified50% confidenceFactExact time
The MoE cache groups expert tensors by data type and shape, allocating one shared slot pool per group in VRAM and using an LRU eviction policy.
1
Sources
50%
Confidence
Long-term
Relevance
10/18/2026
First Seen
Sources
Related Entities
Related Claims
VerifiedThe implementation of the semantic memory layer typically relies on vector databases (such as Chroma, Qdrant, and Pinecone) for vectorized storage and similarity retrieval, a technical approach known as RAG.63% similarUnverifiedLlama 3系列采用了GQA(Grouped Query Attention)设计以压缩KV Cache内存占用62% similarUnverified语义记忆的技术实现依赖向量数据库，本地部署方案包括ChromaDB、Qdrant和FAISS62% similarUnverifiedGQA reduces the size of the KV Cache by allowing multiple Query heads to share the same set of Key-Value heads, which is particularly critical for WebLLM operating in memory-constrained environments.62% similarUnverifiedvLLM's prefix caching implementation is based on a radix tree data structure for fine-grained cache hit matching.61% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1060505API
curl https://kongchang.com/api/v1/knowledge/claims/1060505MCP
get_claim(id=1060505)