Questioning Simplicity Bias Assumptions
Keywords: Simplicity Bias, Shortcut Learning
TL;DR: In this paper we review recent work on the simplicity bias and challenge many of the commonly made assumptions.
Abstract: The Simplicity Bias (SB) is the observation that the training of most commonly used neural network architectures with standard training techniques is biased toward learning simple functions. This phenomenon can be a benefit or drawback depending on the relative complexity of the desired function to be learnt. If the desired function is relatively simple it's a positive. However, if there are simpler features that are highly predictive; commonly named shortcuts or spurious features, that are not present in the test environment, the SB can result in poor generalisation performance. Most existing works on mitigating the SB make various assumptions, either about the features present in the train and test domains or by assuming access to information about the test domain at train time. In this paper we review recent work on the SB and take a critical look at these assumptions.
Primary Area: learning theory
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.
Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 2639
Loading