{"slug": "the-jepa-predictor-a-transferable-operator-for-occluded-feature-completion", "title": "The JEPA Predictor: A Transferable Operator for Occluded Feature Completion", "summary": "Researchers show that the frozen predictor from Joint-Embedding Predictive Architectures (JEPAs) can be transferred across encoder families to improve occluded feature completion. In experiments, pairing CLIP with the I-JEPA predictor lifted fine-grained Stanford Dogs accuracy from 15.9% to 52.1% (+36 percentage points) at heavy occlusion, using only 500 ImageNet-1k images to fit a linear projection between feature spaces. The portable operator requires no retraining of either model and provides a benefit that grows monotonically with mask fraction.", "body_md": "arXiv:2607.16274v1 Announce Type: new\nAbstract: Joint-Embedding Predictive Architectures (JEPAs) train a predictor jointly with their encoder, but downstream deployment discards the predictor and reads features from the encoder alone. The predictor is, by construction, a learned operator from visible-context features to features at masked positions, the structure a partial-view classifier needs. We show that this operator is portable across encoder families. We first establish that, at heavy mask, retaining the frozen predictor on a JEPA encoder substantially closes the accuracy gap against the strongest non-JEPA discriminative baselines. We then bolt the frozen predictors of I-JEPA and V-JEPA 2 onto four non-JEPA hosts (CLIP, DINOv3, DINOv2, MAE) through a single linear projection between feature spaces, fit in closed form on 500 ImageNet-1k images. Across both ImageNet-9 and Stanford Dogs and across three mask fractions, the lift over each host's masked-encoder baseline grows monotonically with the mask fraction K in every host-donor pair. CLIP paired with the I-JEPA predictor recovers most of the accuracy that masking removed on ImageNet-9 at heavy occlusion, and lifts fine-grained Stanford Dogs from 15.9% to 52.1% (+36 pp). The mechanism is identifiable: the projection pays a fixed cost on visible patches and the predictor provides a growing benefit on masked patches; the benefit dominates the heavy-occlusion regime. At low K on fine-grained classification the projection cost exceeds the benefit, defining the boundary where the linear bridge breaks down. The frozen JEPA predictor functions as a portable operator for occluded feature completion across encoder families, requiring no retraining of either model while fitting matched linear probes per mask fraction.", "url": "https://wpnews.pro/news/the-jepa-predictor-a-transferable-operator-for-occluded-feature-completion", "canonical_source": "https://arxiv.org/abs/2607.16274", "published_at": "2026-07-21 04:00:00+00:00", "updated_at": "2026-07-21 04:08:23.444634+00:00", "lang": "en", "topics": ["artificial-intelligence", "machine-learning", "computer-vision", "large-language-models"], "entities": ["Joint-Embedding Predictive Architectures", "I-JEPA", "V-JEPA", "CLIP", "DINOv3", "DINOv2", "MAE", "ImageNet-1k"], "alternates": {"html": "https://wpnews.pro/news/the-jepa-predictor-a-transferable-operator-for-occluded-feature-completion", "markdown": "https://wpnews.pro/news/the-jepa-predictor-a-transferable-operator-for-occluded-feature-completion.md", "text": "https://wpnews.pro/news/the-jepa-predictor-a-transferable-operator-for-occluded-feature-completion.txt", "jsonld": "https://wpnews.pro/news/the-jepa-predictor-a-transferable-operator-for-occluded-feature-completion.jsonld"}}