ProDG: Prototypes for Data-Free Generative Post-Hoc Explainability

TL;DR

ProDG introduces a generative model-based framework that synthesizes prototypes directly from a trained model's weights, enabling data-free, post-hoc interpretability especially useful in privacy-sensitive contexts.

Contribution

It presents the first data-free method for generating prototypes for explainability directly from model weights, removing the need for external datasets.

Findings

Enables high-fidelity prototype generation without data access

Improves interpretability in privacy-sensitive domains

Outperforms existing data-dependent prototype methods

Abstract

Ante-hoc interpretability methods based on prototypes provide highly accurate explanations by utilizing the intuitive "this looks like that" reasoning paradigm. On the other hand, post-hoc models can explain predictions for a single image without relying on an underlying dataset or requiring costly neural network retraining. Recent approaches successfully solve the retraining problem for prototype-based networks. However, they still face a fundamental limitation: they require access to a subset of data (e.g., a test or validation set) to search for and extract the visual prototypes. In this paper, we address this issue and introduce ProDG: Generative Prototypes for Data-Free Post-Hoc Explainability, a novel framework that leverages generative models to synthesize pure, high-fidelity prototypes directly from the frozen model's weights, completely eliminating the dependency on any external data. By establishing this new frontier in Data-Free XAI, ProDG unlocks robust visual interpretability for privacy-sensitive domains, where original data is strictly restricted or fundamentally inaccessible. Project page: https://github.com/piotr310100/ProDG