Diffusion Integrated Gradients: Controllable Path Generation for Flexible Feature Attribution
About
Path-based attribution methods such as Integrated Gradients (IG) are widely adopted for their strong axiomatic properties and effectiveness in attributing model predictions to input features by integrating gradients along a path from a baseline to the input. However, the choice of the attribution path largely affects the quality of explanations, and existing approaches rely on fixed or hand-crafted paths that often produce noisy or distorted attributions. To address this limitation, we propose Diffusion Integrated Gradients (DiffIG), a novel method that reformulates path generation as a conditional generative modeling problem. DiffIG first trains a diffusion model to learn a distribution over paths generated from a Stick-Breaking Process, then employs guided sampling to embed user guidance during the sampling procedure. We demonstrate that DiffIG quantitatively matches or outperforms existing path-based methods, achieving perceptually aligned explanations. This work introduces a new generative perspective for flexible, inference-time controllable Explainable Artificial Intelligence (XAI) methods.
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Feature Attribution | Oxford-IIIT Pet full (val) | DiffID63.68 | 24 | |
| Feature Attribution | Mini-ImageNet 500 randomly sampled images (val) | DiffID Score42.98 | 24 | |
| Attribution Faithfulness | Oxford-IIIT Pet (test) | DiffID57.11 | 8 | |
| Model Attribution | Mini-ImageNet | DiffID0.4298 | 6 |