Patch-Prompt Aligned Bayesian Prompt Tuning for Vision-Language Models

TL;DR

This paper proposes a Bayesian prompt tuning method for vision-language models that generates label-specific stochastic prompts, improving diversity and generalization across various tasks and datasets.

Contribution

It introduces a hierarchical Bayesian framework with semantic regularization for prompt tuning, enhancing diversity and reducing overfitting in vision-language models.

Findings

Improves few-shot image recognition accuracy

Enhances generalization to new categories and datasets

Demonstrates strong transferability across 15 datasets

Abstract

For downstream applications of vision-language pre-trained models, there has been significant interest in constructing effective prompts. Existing works on prompt engineering, which either require laborious manual designs or optimize the prompt tuning as a point estimation problem, may fail to describe diverse characteristics of categories and limit their applications. We introduce a Bayesian probabilistic resolution to prompt tuning, where the label-specific stochastic prompts are generated hierarchically by first sampling a latent vector from an underlying distribution and then employing a lightweight generative model. Importantly, we semantically regularize the tuning process by minimizing the statistical distance between the visual patches and linguistic prompts, which pushes the stochastic label representations to faithfully capture diverse visual concepts, instead of overfitting the training categories. We evaluate the effectiveness of our approach on four tasks: few-shot image recognition, base-to-new generalization, dataset transfer learning, and domain shifts. Extensive results over 15 datasets show promising transferability and generalization performance of our proposed model, both quantitatively and qualitatively.