Hugging Face researchers propose a new method called PEFT, which they claim outperforms LoRA in some tasks. PEFT uses a different approach to fine-tuning language models by adapting the model's weights and adding a small network on top of it. The authors argue that their method is more efficient and effective than LoRA, especially for larger models. They also provide comparisons with other fine-tuning techniques and discuss the potential applications of PEFT. --- Why it matters: This matters to researchers in AI because it challenges the dominance of LoRA as a fine-tuning technique, potentially leading to new breakthroughs in language model performance. The proposed method could also be applied to other areas where fine-tuning is used. Source: https://huggingface.co/blog/peft-beyond-lora