Unverified50% confidenceFactExact time
LoRA由微软研究院于2021年提出，对权重更新量ΔW进行低秩分解ΔW = A·B，参数量通常可压缩至原来的1/100以下
1
Sources
50%
Confidence
Long-term
Relevance
7/13/2026
First Seen
Sources
BERT新增分类类别：增量微调的三种策略与灾难性遗忘解法
Related Claims
UnverifiedLoRA decomposes the fine-tuning weight change ΔW into the product of two low-rank matrices, with a parameter count of only 0.1%–1% of the original model, achieving a parameter compression ratio as high as 256× when r=8 and d=4096.73% similarUnverifiedLoRA (Low-Rank Adaptation) has become the most popular model fine-tuning solution due to significantly reduced VRAM requirements and computational costs.66% similarUnverifiedLoRA reduces the number of trainable parameters from d² to 2dr, where r is typically 4–64.64% similarUnverifiedLoRA等参数高效微调技术可将显存需求降低60%以上61% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/502642API
curl https://kongchang.com/api/v1/knowledge/claims/502642MCP
get_claim(id=502642)