Does CLIP's Generalization Performance Mainly Stem from High Train-Test Similarity?

arXiv:2310.09562v2 Announce Type: replace-cross
Abstract: Foundation models like CLIP are trained on hundreds of millions of samples and effortlessly generalize to new tasks and inputs. Out of the box, CLIP shows stellar zero-shot and few-shot capabilities on a wide range of out-of-distribution (OOD) benchmarks, which prior works attribute mainly to today's large and comprehensive training dataset (like LAION). However, it is questionable how meaningful terms like out-of-distribution generalization are for CLIP as it seems likely that web-scale datasets like LAION …

abstract arxiv benchmarks box capabilities clip cs.ai cs.cv cs.lg distribution few-shot foundation inputs performance prior samples shows stem tasks test train type zero-shot