27cc3576a6f149e95cf68afc3e25cd6c.zip

Reviewers generally agreed that the method offers superior accuracy and efficiency across multiple tasks, supported by thorough ablation studies on design choices.

Reviewers highlighted that the paper's design choices, specifically "feature sharing," were well-motivated and helped the model stay expressive despite the simplifications. Critical Perspectives

The primary consensus among reviewers is that ZIP significantly reduces the "query cost"—the number of times you have to ask the model for a result—while maintaining or improving accuracy. 27cc3576a6f149e95cf68afc3e25cd6c.zip

It addresses the high query requirements of existing methods by reducing problem dimensionality and using "intrinsic-dimensional gradient clipping."

The community recognized the extensive evaluations showcasing superior accuracy and query efficiency over 13+ tasks. Reviewers generally agreed that the method offers superior

This paper introduces a method called designed to improve how we tune large "black-box" models (like CLIP) when we don't have access to their internal code or gradients. Performance and Efficiency

Reviewers pointed out that the soft prompt reparameterization design choices were thoroughly tested, including detailed ablation studies. It addresses the high query requirements of existing

Reviewers from the research community have shared their direct impressions of the work: