# Negative Scores in Feature Ablation and Choosing Baselines for Non-Image Data

**URL:** <https://discuss.pytorch.org/t/negative-scores-in-feature-ablation-and-choosing-baselines-for-non-image-data/205440>\
**Category:** Captum\
**Created:** [June 27, 2024, 11:07pm UTC](https://discuss.pytorch.org/t/negative-scores-in-feature-ablation-and-choosing-baselines-for-non-image-data/205440 "2024-06-27T23:07:12Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![Samira90](https://discuss.pytorch.org/letter_avatar_proxy/v4/letter/s/a3d4f5/32.png) [@Samira90](https://discuss.pytorch.org/u/Samira90)\
**Post date:** [June 27, 2024, 11:07pm UTC](https://discuss.pytorch.org/t/negative-scores-in-feature-ablation-and-choosing-baselines-for-non-image-data/205440/1 "2024-06-27T23:07:12Z")

</div>

How should we interpret a negative score for feature ablation ? Additionally, what recommendations do you have for selecting a baseline when the data is not an image?