What is Backpropagation?

A mathematical objective that quantifies prediction error during training.

A learned numeric value that scales signals passing through a neural network.

An optimized attention algorithm that reduces memory use and speeds up transformer training and inference.

A training setup where a model improves by generating data through interactions or competitions with copies of itself.

When benchmark test examples or close variants are present in training data, inflating reported performance.

A subword tokenization algorithm that merges the most frequent character pairs into reusable tokens.