Unverified50% confidenceFactExact time
During LoRA fine-tuning, the base model remains frozen at all times and is only read, never written to.
1
Sources
50%
Confidence
Long-term
Relevance
8/10/2026
First Seen
Sources
Soup CLI：4GB显卡微调80亿参数大模型的开源工具详解
Related Claims
UnverifiedThe GLM series uses an Autoregressive Blank Infilling training objective, which randomly masks text spans and requires the model to autoregressively restore them58% similarUnverifiedLLM-Wiki's raw data layer follows a read-only strategy, where content in the directory can only be read by the large model and cannot be written to or modified.57% similarUnverifiedA model's knowledge is frozen in the parameter weights at training time and cannot perceive the current time or access real-time data; this is known as the knowledge cutoff problem.52% similarUnverifiedRAG has two vulnerabilities: the model may still misread or misuse the context when processing it, and the retrieved context itself may be incorrect or irrelevant.51% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/726753API
curl https://kongchang.com/api/v1/knowledge/claims/726753MCP
get_claim(id=726753)