Concept
Contrastive objective
A contrastive objective trains a model to pull matching pairs together in an embedding space while pushing mismatched pairs apart. Instead of predicting an exact target (a word, a label, a pixel value), you give the model a set of candidates and ask which ones belong…
The rest of “Contrastive 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→