ReFT: Representation Finetuning for Language Models
Paper
•
2404.03592
•
Published
•
91
In this paper, we propose a strong alternative to PEFTs, LoReFT. LoReFT achieves strong per- formance across benchmarks from four domains while being