# Investigating Saturation Effects in Integrated Gradients

arXiv: https://arxiv.org/abs/2010.12697 · HTML: https://mlsanity.com/paper/2010.12697/

## Key facts

- Cited by 22 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 23 Oct 2020.

## Built on by (most cited first, 22 of 22)

- [Explainability for Large Language Models: A Survey](https://mlsanity.com/paper/2309.01029/) (2023)
- [Guided Integrated Gradients: An Adaptive Path Method for Removing Noise](https://mlsanity.com/paper/2106.09788/) (2021)
- [Discretized Integrated Gradients for Explaining Language Models](https://mlsanity.com/paper/2108.13654/) (2021)
- [Neuron-Level Knowledge Attribution in Large Language Models](https://mlsanity.com/paper/2312.12141/) (2023)
- [Gradient based Feature Attribution in Explainable AI: A Technical Review](https://mlsanity.com/paper/2403.10415/) (2024)
- [Manifold Integrated Gradients: Riemannian Geometry for Feature Attribution](https://mlsanity.com/paper/2405.09800/) (2024)
- [Focus! Rating XAI Methods and Finding Biases](https://mlsanity.com/paper/2109.15035/) (2021)
- [Knowledge Localization: Mission Not Accomplished? Enter Query Localization!](https://mlsanity.com/paper/2405.14117/) (2024)
- [IG2: Integrated Gradient on Iterative Gradient Path for Feature Attribution](https://mlsanity.com/paper/2406.10852/) (2024)
- [Visual-TCAV: Concept-based Attribution and Saliency Maps for Post-hoc Explainability in Image Classification](https://mlsanity.com/paper/2411.05698/) (2024)
- [Probing Deep into Temporal Profile Makes the Infrared Small Target Detector Much Better](https://mlsanity.com/paper/2506.12766/) (2025)
- [Efficient Explanations from Empirical Explainers](https://mlsanity.com/paper/2103.15429/) (2021)
- [Integrated Decision Gradients: Compute Your Attributions Where the Model Makes Its Decision](https://mlsanity.com/paper/2305.20052/) (2023)
- [Negative Flux Aggregation to Estimate Feature Attributions](https://mlsanity.com/paper/2301.06989/) (2023)
- [Saliency strikes back: How filtering out high frequencies improves white-box explanations](https://mlsanity.com/paper/2307.09591/) (2023)
- [Applications of interpretable deep learning in neuroimaging: a comprehensive review](https://mlsanity.com/paper/2406.17792/) (2024)
- [Leaping Into Memories: Space-Time Deep Feature Synthesis](https://mlsanity.com/paper/2303.09941/) (2023)
- [Derivative-Free Diffusion Manifold-Constrained Gradient for Unified XAI](https://mlsanity.com/paper/2411.15265/) (2024)
- [CIDR: A Cooperative Integrated Dynamic Refining Method for Minimal Feature Removal Problem](https://mlsanity.com/paper/2312.08157/) (2023)
- [PE: A Poincare Explanation Method for Fast Text Hierarchy Generation](https://mlsanity.com/paper/2403.16554/) (2024)
- [Feature Attribution Stability Suite: How Stable Are Post-Hoc Attributions?](https://mlsanity.com/paper/2604.02532/) (2026)
- [WOMBAT: Whitebox Oracle for Molecular Benchmarking and Attribution Testing](https://mlsanity.com/paper/2610.00713/) (2026)

## Abstract (arXiv)

Integrated Gradients has become a popular method for post-hoc model interpretability. De-spite its popularity, the composition and relative impact of different regions of the integral path are not well understood. We explore these effects and find that gradients in saturated regions of this path, where model output changes minimally, contribute disproportionately to the computed attribution. We propose a variant of IntegratedGradients which primarily captures gradients in unsaturated regions and evaluate this method on ImageNet classification networks. We find that this attribution technique shows higher model faithfulness and lower sensitivity to noise com-pared with standard Integrated Gradients. A note-book illustrating our computations and results is available at https://github.com/vivekmig/captum-1/tree/ExpandedIG.