Concept
Task-completion objective
A task-completion objective is a training loss computed only over the tokens the model is supposed to produce, not over the tokens it was given. In supervised instruction tuning, each example is an (instruction, response) pair, and the loss is masked so that only the…
The rest of “Task-completion objective” is a premium feature: every concept in the library gets a precise, practitioner-focused write-up like this one, cross-linked straight from the paper summaries that use it.
Log in to unlock→