With the core motivation of VCC providing reproducible experiments, we will require participants to submit a system description report at the end of the challenge.
The report should disclose the details of the system (features, model architecture, training and inference details, etc.), the datasets used, and the results.
More instructions will be shared later.
For this year, we will be allowing the use of additional datasets, including internal ones.
Any dataset except for GTSinger is allowed for training. Participants should also ensure that any pretrained models they use are not trained on GTSinger.
However, for the sake of reproducibility, we will clarify in the final overview paper which systems used internal and purely open-sourced datasets. Thus, we will require participants to disclose which datasets they used in their systems in the system description report.
We are planning to submit a proposal for a workshop paper with the final results in an AAAI 2026 workshop.
Participants are encouraged to submit a paper describing their entry. More details will be shared later.
svcc2025__at__vc-challenge.org
Back to Challenge Homepage