Unverified50% confidenceFactExact time
LoRA was proposed by Hu et al. in 2021; it freezes the original weights and injects two low-rank matrices A and B for parameter-efficient fine-tuning, typically reducing the number of trainable parameters by more than 99%.
1
Sources
50%
Confidence
Long-term
Relevance
9/18/2026
First Seen
Sources
Related Entities
Related Claims
VerifiedLoRA (Low-Rank Adaptation) was proposed by Microsoft in 2021, with the core idea of freezing the weights of the pretrained model and only training injected low-rank decomposition matrices.77% similarVerifiedLoRA reduces fine-tuning costs by training only a small number of adapter layers (low-rank matrix bypass training).76% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/931302API
curl https://kongchang.com/api/v1/knowledge/claims/931302MCP
get_claim(id=931302)