Unverified50% confidenceFactExact time
ZHA uses distillation to give an already frozen LLM sparse attention capability without the need to retrain the base model.
1
Sources
50%
Confidence
Long-term
Relevance
11/3/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedZHA's contribution may lie in combining tree-index sparse attention with distillation of a frozen model to form a reusable retrofit pipeline.71% similarUnverified当前的LLM一旦完成训练其参数即被冻结，只能通过微调或RAG等外部机制来更新70% similarUnverifiedLLM在推理阶段参数是冻结的，模型不会因用户交互而更新权重，因此本质上是无状态的69% similarUnverifiedLLM采用无状态设计，模型权重在训练后被固化，推理时不会因某次对话而改变内部参数68% similarUnverifiedThe current mainstream LLM applications adopt a stateless interaction model, where the model itself does not retain cross-session memory and all context must be temporarily fed in via prompts.66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1129241API
curl https://kongchang.com/api/v1/knowledge/claims/1129241MCP
get_claim(id=1129241)