Causality-aligned Prompt Learning via Diffusion-based Counterfactual Generation
This addresses the challenge of robust feature generalization across categories in prompt learning, representing a novel method for a known bottleneck rather than an incremental improvement.
The paper tackles the problem of achieving causally invariant prompts in prompt learning by introducing DiCap, a diffusion-based counterfactual prompt learning framework that generates counterfactuals satisfying minimal sufficiency criteria, resulting in excellent performance across tasks like image classification, image-text retrieval, and visual question answering with strong advantages in unseen categories.
Prompt learning has garnered attention for its efficiency over traditional model training and fine-tuning. However, existing methods, constrained by inadequate theoretical foundations, encounter difficulties in achieving causally invariant prompts, ultimately falling short of capturing robust features that generalize effectively across categories. To address these challenges, we introduce the $\textit{\textbf{DiCap}}$ model, a theoretically grounded $\textbf{Di}$ffusion-based $\textbf{C}$ounterf$\textbf{a}$ctual $\textbf{p}$rompt learning framework, which leverages a diffusion process to iteratively sample gradients from the marginal and conditional distributions of the causal model, guiding the generation of counterfactuals that satisfy the minimal sufficiency criterion. Grounded in rigorous theoretical derivations, this approach guarantees the identifiability of counterfactual outcomes while imposing strict bounds on estimation errors. We further employ a contrastive learning framework that leverages the generated counterfactuals, thereby enabling the refined extraction of prompts that are precisely aligned with the causal features of the data. Extensive experimental results demonstrate that our method performs excellently across tasks such as image classification, image-text retrieval, and visual question answering, with particularly strong advantages in unseen categories.