IC-147CLIP's OOD performance on rendition domains is largely an artifact of domain contamination in its web-scale training data

Prasanna Mayilvahanan, Roland S. Zimmermann, Thaddäus Wiedemer, Evgenia Rusak, Attila Juhos, Matthias Bethge, Wieland Brendel

SourceIn Search of Forgotten Domain Generalization

The paper demonstrates that CLIP's strong zero-shot accuracy on rendition-style test sets (ImageNet-Sketch, ImageNet-R, DomainNet variants) is substantially inflated by the presence of rendition images in its LAION-400M/200M training data. By training CLIP ViT-B/32 from scratch on a clean 57M-sample natural-image-only subset (LAION-natural) and comparing to equally-sized subsets of the contaminated LAION-200M, the authors show that relative corrected OOD accuracy on rendition test sets drops to approximately 0.4. Adding just 13–16M rendition images to the training set nearly doubles rendition top-1 accuracy (from ~22% to ~40–42%) while having minimal impact on natural-domain performance. The effect is consistent across training set sizes from 4M to 57M samples, and the authors conclude that CLIP's apparent OOD robustness is not genuine generalization but a consequence of overlapping domain data.

Evidence
correlational
Key metric
Relative corrected OOD accuracy ~0.4 on rendition test sets (natural-only training); LAION-natural 21.98% rendition top-1 (standard) / 17.81% (clean); LAION-mix-13m 40.48% / 40.78%; LAION-mix-16m 41.46% / 42.07%; LAION-rand-57m 40.66% / 39.58%; LAION-200M contains at least 7.90% rendition images (16M samples)
Caveat
Training sets are constrained by the availability of natural (57M) and rendition (16M) samples; the authors conjecture but cannot prove the result holds at the full LAION-400M scale. They also note they did not control for object class distribution differences between LAION-natural and LAION-rendition, and the impact of ambiguous samples is left for future work.
Model
CLIP / CLIP-ViT (LC)
Concepts
Shortcut, Method artefact
Datasets
LAION-200M [source], ImageNet-R / ImageNet-Rendition [eval], ImageNet-Sketch [eval], DomainNet [eval], ImageNet-A [eval], ObjectNet [eval], ImageNet-V2 [eval]
Methods
OpenCLIP [primary]
Related work
Mayilvahanan et al. 2023 (Does CLIP's generalization performance mainly stem from high train-test similarity?) [compared-to]
Extraction
automatic-extraction