"Oops, I Erred" — Collecting, Curating and Using Imperfect Data for Realistic Scenarios

Call for Contributions

We solicit two kinds of contributions: papers and datasets.

Papers

Papers may be short (4 pages) or long (8 pages) and cover any aspect of collecting, curating, or using imperfect data, including position papers. All submissions are reviewed double-blind on OpenReview. Following CoRL policy, we will not accept work already accepted to the main conference.

Datasets

Datasets are solicited as first-class contributions. Each dataset upload must be accompanied by a short white-paper (max 4 pages) describing the data's origin, use-case, and scope, with documentation that allows users to use the data. We recommend following the Oopsie formatting standards for contributors submitting manipulation data, or to define similar standards for other types of embodiments and tasks. Contributors present via a poster, video, or a live artifact demo.

Datasets following the Oopsie-Data requirements and format can voluntarily be included in the Oopsie-Data project as well.

To reward the often-invisible labor of dataset creation, we extend award mechanisms normally reserved for papers to datasets: participants and panelists vote for their favorite data contributions during the poster session. We will define a few criteria for dataset awards soon, stay tuned!

Topics of Interest

• Collecting robotic datasets of failures, near-misses, and out-of-distribution events at scale
• Classifying, annotating, and formatting imperfect data for broad usability
• Using imperfect data to improve policy learning, world models, or reward models
• Data sharing infrastructure, incentive structures, and standardization
• Position papers on the value (or lack thereof) of particular forms of imperfect data

Important Dates

Review and Selection Process

• All reviews will be double-blind
• All accepted papers will have posters at the workshop
• Papers will have non-archival status
• At least one co-author of an accepted paper is expected to attend the workshop
• All accepted papers will be made public on OpenReview