Gram-Anchored Prompt Learning for Vision-Language Models via Second-Order Statistics

1. 1Identify the limitations of first-order visual features in prompt learning
2. 2Incorporate second-order statistics to capture spatially disentangled feature correlations
3. 3Implement Gram-Anchored Prompt Learning to adapt Vision-Language Models to downstream tasks

ML researchers and engineers working on Vision-Language Models can benefit from this approach to improve model adaptation and robustness

💡 Incorporating second-order statistics can enhance robust adaptation of Vision-Language Models

💡 Improve VLMs with Gram-Anchored Prompt Learning via second-order statistics!