AdaptLLM/law_knowledge_prob

Adapting LLMs to Domains via Continual Pre-Training (ICLR 2024) This repo contains the Law Knowledge Probing dataset used in our paper Adapting Large Language Models via Reading Comprehension. We explore continued pre-training on domain-specific corpora for large language models. While this approach enriches LLMs with domain knowledge, it significantly hurts their prompting ability for question answering. Inspired by human learning via reading comprehension, we propose a simple… See the full description on the dataset page: https://huggingface.co/datasets/AdaptLLM/law_knowledge_prob.