Solicited Topics
We invite submissions that address discovery, learning, and unification in the presence of spurious correlations. We welcome a wide range of topics, including but not limited to:
Methods for discovering and diagnosing spurious correlations.
Evaluation and stress tests of model stability.
Impacts of different dataset shifts when learning exploits a shortcut/spurious correlation.
Learning robust models in the presence of spurious correlations.
Exploring relationships b/n methods from causal ML, algorithmic fairness, and OOD generalization.
Furthermore, we strongly encourage practitioners to submit examples of failure modes due to spurious correlations in real-world scenarios. We are particularly interested in submissions that can create new opportunities for collaboration, and motivate foundational research that is impactful in real-world applications.
Important dates and instructions
- Submission deadline: May 30 (Anywhere on Earth)
- Decision notification: June 19th (Anywhere on Earth)
- Workshop date: Saturday, July 29
Instructions
Submission URL: https://openreview.net/group?id=ICML.cc/2023/Workshop/SCIS
Formatting Instructions
All submissions must be in PDF format. Submissions can have up to four content pages excluding references and appendices. Additional supplementary materials (e.g., appendices) can be submitted with the main manuscript. Reviewers will only be required to read the main paper. Please use the LaTeX style files.
Non-archival submissions
Work may be previously published, completed, or ongoing. The workshop will not publish proceedings but all accepted papers will be posted on our workshop website. Accepted papers can be submitted to other venues later.
Double-blind reviews
Submissions will be peer-reviewed by at least 2 reviewers, in addition to an area chair. The reviewing process will be double-blind at the level of the reviewers. As an author, you are responsible for anonymizing your submission. Do not include any authors' names, affiliations, acknowledgements, or any other information that could result in de-anonymization.