MJEPA: A Simple and Scalable Joint-Embedding Predictive Architecture for Audio-Visual Learning
Quick Answer
MJEPA introduces a unified joint-embedding predictive architecture for audio-visual learning, outperforming prior models by over 6.8 mAP on AudioSet-20K.
Quick Take
MJEPA introduces a unified joint-embedding predictive architecture for audio-visual learning, outperforming prior models by over 6.8 mAP on AudioSet-20K. Utilizing a single predictive objective across modalities, it enhances representation synergy while using 10x less video data, demonstrating significant efficiency and effectiveness.
Key Points
- MJEPA uses a single unified encoder for both audio and visual modalities.
- Cross-modal prediction is essential for improving representation quality.
- Outperformed previous frozen baselines by over 6.8 mAP on AudioSet-20K.
- Achieved competitive results on ESC-50 and FSD50K datasets.
- Utilizes 10x less video data compared to fully finetuned models.
Paper Resources
📖 Reader Mode
~2 min readAbstract:Self-supervised learning from large-scale video data has emerged as a dominant paradigm for visual representation learning. Since audio and visual streams naturally co-occur in video data, extending this success to jointly learn from both modalities is a natural next step, yet it remains challenging. Existing audio-visual self-supervised methods rely on modality-specific encoders and complex combinations of contrastive or reconstruction objectives, limiting cross-modal synergy and scalability. Joint Embedding Predictive Architectures (JEPAs) offer a simple, modality-agnostic alternative, but have to date been applied primarily to individual modalities. We introduce MJEPA, a joint-embedding predictive architecture for audio-visual learning that uses a single, unified encoder for both modalities. Our approach uses only a single predictive objective, applied both within and across modalities. We show that cross-modal prediction is critical: without it, a shared encoder degrades below unimodal baselines; with it, each modality's representation benefits from the other. Our frozen ViT-g model outperforms the best prior frozen baseline by over 6.8 mAP on AudioSet-20K, surpasses fully finetuned models on ESC-50 and FSD50K, and is competitive on video benchmarks despite using 10x less video data.
| Subjects: | Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG) |
| Cite as: | arXiv:2606.25225 [cs.CV] |
| (or arXiv:2606.25225v1 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2606.25225 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Revant Teotia [view email]
[v1]
Tue, 23 Jun 2026 22:48:42 UTC (6,144 KB)
— Originally published at arxiv.org
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.CV
See more →LLM-Guided ANN Index Optimization for Human-Object Interaction Retrieval
A phase-aware LLM agent optimizes human-object interaction retrieval, outperforming Optuna TPE by 33.3% and VDTuner by 34.2% on the HICO-DET benchmark. This method enhances throughput by 15.3x over UniIR and demonstrates strong transferability across vector database management systems.