Unverified50% confidenceFactExact time

LoRA将可训练参数量从d²降低到2dr，其中r通常为4-64

1

Sources

50%

Confidence

Medium-term (~90 days)

Relevance

7/2/2026

First Seen

Valid until: 9/30/2026

Sources

AI大模型学习路线：零基础入门到项目实战完整指南

bilibiliAI-大模型应用开发6/5/2026

Related Claims

VerifiedLoRA中低秩矩阵的秩r通常设为8或16，可训练参数量从d×k降至r×(d+k)，降幅可达数百倍78% similar UnverifiedLoRA可以将可训练参数量减少99%以上75% similar UnverifiedLoRA将微调权重变化量ΔW分解为两个低秩矩阵的乘积，参数量仅为原模型的0.1%~1%，当r=8、d=4096时参数压缩比高达256倍69% similar Partially Verified对于一个4096×4096的权重矩阵，LoRA可能只需要训练两个4096×8的矩阵，将可训练参数量从约1600万降低到约6.5万，减少了99.6%68% similar

Cite This Claim

Stable URI

https://kongchang.com/claim/53398

API

curl https://kongchang.com/api/v1/knowledge/claims/53398

MCP

get_claim(id=53398)