ConceptAI & ML
Scaling laws for neural language models
The empirical finding that cross-entropy loss improves predictably as a power law with compute, dataset size, and model size.
Know first
Named as a prerequisite by the lectures that define this. The number is how many say so. A grey one has no page yet.
Compute budget1Language modeling loss1
Lecturers
1Connections
1The lectures that define it or assume it, and the concepts next to it.