What is Confidence Interval?

A language-model metric measuring how surprised the model is by true next tokens.

How well a model's confidence scores match actual correctness probabilities.

The processing resources required to train and run models, often measured in FLOPS or GPU hours.

An incorrect prediction where a model misses a true positive case.

A configuration value set before training, such as learning rate, batch size, or depth.

A decoding strategy that samples only from the k most likely next tokens.