Verified65% confidenceFactExact time
In LoRA, matrix A is initialized with random Gaussian values and matrix B is initialized to zero, ensuring that the added quantity is zero at the start of training so as not to damage the original model's capabilities, and introducing no additional latency during inference.
3
Sources
65%
Confidence
Long-term
Relevance
7/24/2026
First Seen
Sources
零基础入门大模型微调：原理、场景与实战路径
Related Claims
UnverifiedTraining from scratch means not using any pretrained weights, but instead completing the full training process on a new dataset with randomly initialized parameters.78% similarVerifiedAfter LoRA training is complete, the low-rank matrices can be merged back into the original weights, introducing no additional latency during inference.75% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/610905API
curl https://kongchang.com/api/v1/knowledge/claims/610905MCP
get_claim(id=610905)