Tag
This paper investigates instruction finetuning of DeepSeek-R1-8B using LoRA and NEFTune for financial named-entity recognition, achieving a micro-F1 of 0.912 and outperforming several baseline models.