Follow
Beyond Scale: the Diversity Coefficient as a Data Quality Metric Demonstrates LLMs are Pre-trained on Formally Diverse Data. (arXiv:2306.13840v1 [cs.CL])
Beyond Scale: the Diversity Coefficient as a Data Quality Metric Demonstrates LLMs are Pre-trained on Formally Diverse Data. (arXiv:2306.13840v1 [cs.CL])