Leveraging Diffusion Disentangled Representations to Mitigate Shortcuts in Underspecified Visual Tasks
This addresses the problem of spurious correlations in computer vision for researchers, though it is incremental as it builds on existing ensemble and diffusion model techniques.
The paper tackles shortcut learning in underspecified visual tasks by proposing an ensemble diversification framework that uses diffusion models to generate synthetic counterfactuals, achieving ensemble diversity performance comparable to methods needing extra data.
Spurious correlations in the data, where multiple cues are predictive of the target labels, often lead to shortcut learning phenomena, where a model may rely on erroneous, easy-to-learn, cues while ignoring reliable ones. In this work, we propose an ensemble diversification framework exploiting the generation of synthetic counterfactuals using Diffusion Probabilistic Models (DPMs). We discover that DPMs have the inherent capability to represent multiple visual cues independently, even when they are largely correlated in the training data. We leverage this characteristic to encourage model diversity and empirically show the efficacy of the approach with respect to several diversification objectives. We show that diffusion-guided diversification can lead models to avert attention from shortcut cues, achieving ensemble diversity performance comparable to previous methods requiring additional data collection.