Code
Beyond Scaling: Measuring and Predicting the Upper Bound of Knowledge Retention in Language Model Pre-Training
Why Lacuna links this resource
Confirmed by the paper and resource
The paper explicitly states that its dataset and code are available at the provided GitHub URL. The repository's description and README explicitly state it is for the paper with the exact same title and link directly to the paper's arXiv ID, confirming it is the primary resource released by the authors.