GIFT-SW: Gaussian noise Injected Fine-Tuning of Salient Weights for LLMs

提交: 2024-09-25更新: 2024-10-15

TL;DR

Fine-tuning weights in salient columns produces results comparable to those obtained using low-rank adapters

关键词

large language modelsparameter efficient fine-tuningGaussian noisequantization

评审与讨论

编辑台拒稿- Submission Desk Rejected by Program Chairs

2024-10-15

material on 11th page, not reproducibility or ethics statement