Unverified50% confidenceFactExact time
In LoRA's scaling factor α/r, dividing by r serves a normalization purpose, so that adjusting r does not require a simultaneous adjustment of the learning rate.
1
Sources
50%
Confidence
Long-term
Relevance
9/10/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedLoRA推理时可将BA与原权重合并，不增加任何推理延迟，通常只应用于注意力层的Q、V权重矩阵68% similarVerifiedLoRA是以参数高效微调（PEFT）为代表的主流微调方法，核心思想是在不改动原始模型权重的前提下向特定层注入低秩矩阵，只训练新增参数66% similarVerifiedLoRA achieves fine-tuning by inserting low-rank decomposition matrices alongside the model's weight matrices, typically requiring training of only 0.1%–1% of the original parameter count.65% similarUnverified微调通过在私有数据上继续训练来更新模型权重，RAG则在不修改模型的前提下通过检索外挂知识库增强回答63% similarVerifiedWith LoRA, the weights can be merged back into the original model at inference time, introducing no additional latency63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/888472API
curl https://kongchang.com/api/v1/knowledge/claims/888472MCP
get_claim(id=888472)