Call For Papers
Beyond our Vibe Reproducibility challenge, we also welcome authors to document experiences in attempting to reproduce papers presented at previous AAAI conferences (Main Track) and papers that have been published in the Journal of Artificial Intelligence Research (JAIR), or other academic AI/ML venues. Such works are most influential and useful when they document insights and difficulties in reproduction, what worked well, and distil their knowledge into a realization that may be applicable to other papers in the future.
More broadly, we welcome submissions of both full technical papers and position papers that conduct novel and unpublished research on reproducibility in AI. Papers should be written in English and formatted according to the AAAI style. Full papers should not exceed eight pages, while position papers must not exceed five pages (unlimited references in both cases). Position papers will be presented alongside regular papers by the authors. Selected papers will be invited for an oral or poster presentation. Any technical study of reproducibility is within scope for our workshop, as we understand data from other fields/domains may be advantageous in its availability to study questions about reproducibility. As examples, topics in this area may seek to understand the science behind:
- How often code and papers are consistent in what is described
- How citations or other incentives relate to reproducibility
- How different low-precision training regimes or hardware can impact replication success
- Or many more other possible topics!
Dates
Submission Deadline: October 22, AOE
Notification date: November 5