Contrastive Auxiliary Training in Video-JEPA for Generalization to Unseen Actions in Few-Shot Settings on EPIC-Kitchens-100

Joint-Embedding Predictive Architectures (JEPA) are a promising framework for self-supervised video representation learning, yet the behavior of auxiliary objectives in small-scale Video-JEPA training is not well characterized. We report a small-scale empirical study of 18 auxiliary objective variants for Video-JEPA across two pretraining regimes: single-dataset (UCF-101) and mixed-dataset (UCF-101 + Something-Something V2 + ImageNet-100). We evaluate frozen representations on three complementary benchmarks: Diving-48 (fine-grained motion), SomethingSomething V2 (temporal reasoning), and Image

Research goal: Does contrastive auxiliary training in Video-JEPA improve generalization to unseen actions in few-shot settings when evaluated on the EPIC-Kitchens-100 dataset compared to reconstructive objectives?

Autonomous synthesis report generated by SOVEREIGN Research Kernel. Tribunal consensus score: 7.5/10.