Instruction Finetuning DeepSeek-R1-8B Model Using LoRA and NEFTune

In three linesDeepSeek-R1-8B fine-tuned with LoRA and NEFTune for financial named-entity recognition. On 1693 annotated samples, the model achieves micro-F1 of 0.912 across 7 entity types, outperforming Llama3-8B, Qwen3-8B, and Baichuan2-7B.

Summary generated by Claude — human-verified