Hoffmann et al. (2022) attribute to Rae et al. (2021) the finding that improving loss by adding parameters has a negligible effect on unprompted toxic generation, and test whether it holds for compute-optimal training.
One recorded claim. The claim panel gives its qualification, the source passages with their place in each source, and how it was checked.
What it connects
Training Compute-Optimal Large Language Models → attributes to → Scaling Language Models: Methods, Analysis & Insights from Training Gopher
Where it appears
No collection page holds this claim. It is listed with its records in the explorer.