Verified75% confidenceFactExact time
The principle of LoRA is to decompose the weight change ΔW during fine-tuning into the product of two low-rank matrices (ΔW = A×B).
3
Sources
75%
Confidence
Long-term
Relevance
9/10/2026
First Seen
Sources
Related Entities
Related Claims
VerifiedLoRA attaches two low-rank matrices A and B in parallel alongside each target weight matrix, where A is randomly initialized and B is initialized to zero; only these two small matrices are trained, and during inference the product A×B is added back to the original weights.72% 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.72% similarUnverifiedDuring LoRA inference, W' = W₀ + (α/r)BA can be pre-computed to merge the LoRA branch into the original weights, achieving zero additional inference latency.70% similarVerifiedLoRA（低秩适应）的核心思想是模型权重更新矩阵本征秩很低，无需更新全部参数，只需并联两个低秩矩阵67% similarUnverifiedLoRA 由 Hu 等人在 2021 年提出，通过为每层注入可训练的低秩分解矩阵实现参数更新 ΔW = BA，将可训练参数量压缩至原始模型的 0.1%-1%65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/889924API
curl https://kongchang.com/api/v1/knowledge/claims/889924MCP
get_claim(id=889924)