LoRA
LoRA
Why it matters
Freezes the weights and trains only low-rank matrices, cutting trainable parameters to 1/10,000 for GPT-3 175B, with no increase in inference latency
Source — Primary sources first; secondary reports are labelled “Press”
Paperarxiv.orghttps://arxiv.org/abs/2106.09685Added in Week 35, 2026 · See that week