LORE: Lagrangian-Optimized Robust Embeddings for Visual Encoders
–Neural Information Processing Systems
Visual encoders have become fundamental components in modern computer vision pipelines. However, ensuring robustness against adversarial perturbations remains a critical challenge. Recent efforts have explored both supervised and unsupervised adversarial fine-tuning strategies. We identify two key limitations in these approaches: (i) they often suffer from instability, especially during the early stages of fine-tuning, resulting in suboptimal convergence and degraded performance on clean data, and (ii) they exhibit a suboptimal trade-off between robustness and clean data accuracy, hindering the simultaneous optimization of both objectives. To overcome these challenges, we propose Lagrangian-Optimized Robust Embeddings (LORE), a novel unsupervised adversarial fine-tuning framework.
Neural Information Processing Systems
Jun-18-2026, 14:36:22 GMT
- Country:
- North America > United States (0.27)
- Genre:
- Research Report
- New Finding (1.00)
- Experimental Study (1.00)
- Research Report
- Industry:
- Information Technology (0.46)
- Education (0.45)
- Technology: