A Framework to Implement 1+N Multi-task Fine-tuning Pattern in LLMs Using the CGC-LORA Algorithm

With the productive evolution of large language models (LLMs) in the field of natural language processing (NLP), tons of effort has been made to effectively fine-tune common pre-trained LLMs to fulfill a variety of tasks in one or multiple specific domain. In practice, there are two prevailing ways, in which the adaptation can be achieved: (i) Multiple Independent Models: Pre-trained LLMs are fine-tuned a few times independently using the corresponding training samples from each task. (ii) An Integrated Model: Samples …

algorithm cs.ai cs.cl cs.lg domain evolution fine-tuning framework language language models language processing large language large language models llms lora multiple natural natural language natural language processing nlp practice processing productive tasks