Rethinking JEPA: Compute-Efficient Video SSL with Frozen Teachers

• URL: http://arxiv.org/abs/2509.24317v1
• Date: Mon, 29 Sep 2025 05:55:17 GMT
• Title: Rethinking JEPA: Compute-Efficient Video SSL with Frozen Teachers
• Authors: Xianhang Li, Chen Huang, Chun-Liang Li, Eran Malach, Josh Susskind, Vimal Thilak, Etai Littwin,
• Abstract summary: Video Joint Embedding Predictive Architectures (V-JEPA) learn generalizable off-the-shelf video representation by predicting masked regions in latent space with an exponential moving average (EMA)-updated teacher.<n>We revisit masked-latent prediction and show that a frozen teacher suffices.<n>We find that student quality is remarkably robust to teacher quality: high-performing students emerge even with small, sub-optimal teachers.
• Score: 37.91964614027141
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Video Joint Embedding Predictive Architectures (V-JEPA) learn generalizable off-the-shelf video representation by predicting masked regions in latent space with an exponential moving average (EMA)-updated teacher. While EMA prevents representation collapse, it complicates scalable model selection and couples teacher and student architectures. We revisit masked-latent prediction and show that a frozen teacher suffices. Concretely, we (i) train a target encoder with a simple pixel-reconstruction objective under V-JEPA masking, then (ii) freeze it and train a student to predict the teacher's latents on masked regions. This leads to a two-stage, unregularized scheme that we refer to as SALT (Static-teacher Asymmetric Latent Training). SALT decouples optimization into pixel reconstruction (teacher) and masked latent prediction (student), increasing transparency, efficiency, and scalability while preserving the ability of representation to generalize under frozen evaluation. Empirically, our student models outperform recently proposed V-JEPA 2 encoders under frozen backbone evaluation across diverse benchmarks. They are also more compute-optimal: at matched pretraining FLOPs, our method achieves higher probing accuracy, and its scaling curves dominate V-JEPA's accuracy-FLOPs Pareto frontier. Finally, we find that student quality is remarkably robust to teacher quality: high-performing students emerge even with small, sub-optimal teachers. This points to a compute budget allocation that should overwhelmingly favor the student. These results position SALT as a simple, scalable, and compute-efficient alternative to EMA-based self-distillation for video representation learning.

Related papers

• US-JEPA: A Joint Embedding Predictive Architecture for Medical Ultrasound [5.216814358105614]
  Ultrasound (US) imaging poses unique challenges for representation learning due to its inherently noisy acquisition process.<n>Joint-Embedding Predictive Architectures (JEPAs) predict masked latent representations rather than raw pixels.<n>We propose US-JEPA, a self-supervised framework that adopts the Static-teacher Asymmetric Latent Training (SALT) objective.
  arXiv  Detail & Related papers  (2026-02-22T19:56:56Z)
• Gap Preserving Distillation by Building Bidirectional Mappings with A Dynamic Teacher [43.678380057638016]
  Gap Preserving Distillation (GPD) method trains an additional dynamic teacher model from scratch along with training the student to bridge this gap. In experiments, GPD significantly outperforms existing distillation methods on top of both CNNs and transformers architectures. GPD also generalizes well to the scenarios without a pre-trained teacher, including training from scratch and fine-tuning, yielding a large improvement of 1.80% and 0.89% on ResNet18.
  arXiv  Detail & Related papers  (2024-10-05T12:29:51Z)
• SIGMA:Sinkhorn-Guided Masked Video Modeling [69.31715194419091]
  Sinkhorn-guided Masked Video Modelling ( SIGMA) is a novel video pretraining method. We distribute features of space-time tubes evenly across a limited number of learnable clusters. Experimental results on ten datasets validate the effectiveness of SIGMA in learning more performant, temporally-aware, and robust video representations.
  arXiv  Detail & Related papers  (2024-07-22T08:04:09Z)
• Skeleton2vec: A Self-supervised Learning Framework with Contextualized Target Representations for Skeleton Sequence [56.092059713922744]
  We show that using high-level contextualized features as prediction targets can achieve superior performance. Specifically, we propose Skeleton2vec, a simple and efficient self-supervised 3D action representation learning framework. Our proposed Skeleton2vec outperforms previous methods and achieves state-of-the-art results.
  arXiv  Detail & Related papers  (2024-01-01T12:08:35Z)
• Improving Knowledge Distillation via Regularizing Feature Norm and Direction [16.98806338782858]
  Knowledge distillation (KD) exploits a large well-trained model (i.e., teacher) to train a small student model on the same dataset for the same task. Treating teacher features as knowledge, prevailing methods of knowledge distillation train student by aligning its features with the teacher's, e.g., by minimizing the KL-divergence between their logits or L2 distance between their intermediate features. While it is natural to believe that better alignment of student features to the teacher better distills teacher knowledge, simply forcing this alignment does not directly contribute to the student's performance, e.g.
  arXiv  Detail & Related papers  (2023-05-26T15:05:19Z)
• EmbedDistill: A Geometric Knowledge Distillation for Information Retrieval [83.79667141681418]
  Large neural models (such as Transformers) achieve state-of-the-art performance for information retrieval (IR) We propose a novel distillation approach that leverages the relative geometry among queries and documents learned by the large teacher model. We show that our approach successfully distills from both dual-encoder (DE) and cross-encoder (CE) teacher models to 1/10th size asymmetric students that can retain 95-97% of the teacher performance.
  arXiv  Detail & Related papers  (2023-01-27T22:04:37Z)
• Masked Video Distillation: Rethinking Masked Feature Modeling for Self-supervised Video Representation Learning [123.63301596019522]
  Masked video distillation (MVD) is a simple yet effective two-stage masked feature modeling framework for video representation learning. For the choice of teacher models, we observe that students taught by video teachers perform better on temporally-heavy video tasks. We design a spatial-temporal co-teaching method for MVD to leverage the advantage of different teachers.
  arXiv  Detail & Related papers  (2022-12-08T18:59:59Z)
• Exploring The Role of Mean Teachers in Self-supervised Masked Auto-Encoders [64.03000385267339]
  Masked image modeling (MIM) has become a popular strategy for self-supervised learning(SSL) of visual representations with Vision Transformers. We present a simple SSL method, the Reconstruction-Consistent Masked Auto-Encoder (RC-MAE) by adding an EMA teacher to MAE. RC-MAE converges faster and requires less memory usage than state-of-the-art self-distillation methods during pre-training.
  arXiv  Detail & Related papers  (2022-10-05T08:08:55Z)
• ALP-KD: Attention-Based Layer Projection for Knowledge Distillation [30.896957367331137]
  Two neural networks, namely a teacher and a student, are coupled together during training. The teacher network is supposed to be a trustworthy predictor and the student tries to mimic its predictions. In such a setting, distillation only happens for final predictions whereas the student could also benefit from teacher's supervision for internal components.
  arXiv  Detail & Related papers  (2020-12-27T22:30:13Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.