Challenge results¶

The main challenge has concluded, and we are proud to recognise the following winning teams:

Test ALL¶

Baseline UNICORN Score: 0.378

🥇 MEVIS - 0.456

🥈 AIMHI-MEDAI - 0.447

🥉 kaiko - 0.273

Test Radiology Vision¶

Baseline Radiology Score: 0.077

🥇 MEVIS - 0.458

🥈 AIMHI-MEDAI - 0.341

🥉 Curia - 0.038

Test Pathology Vision¶

Baseline Pathology Score: 0.334

🥇 MEVIS - 0.482

🥈 AIMHI-MEDAI - 0.356

🥉 AIVIS - 0.348

Test Language¶

Baseline Language Score: 0.629

🥇 AIMHI-MEDAI - 0.622

🥈 Exynos - 0.595

🥉 kaiko - 0.544

Early Bird¶

🥇Early Bird prize to: AIMHI-MEDAI

✨ Congratulations to all participating teams for their outstanding efforts! The final test leaderboards are now also visible, you can view them here.

The UNICORN Score reflects performance across all 20 tasks. If a submission covers only a subset of tasks, the worst possible score is assigned for the tasks not included. This ensures that the UNICORN score remains comparable across all leaderboards. The Radiology, Pathology, and Language scores represent the average performance across tasks within each specific leaderboard. These averages include only the tasks belonging to that leaderboard. The normalization logic used to compute the UNICORN Score is available in our unicorn_eval repository here.

We would like to thank our official sponsors — ImFusion GmbH, ScreenPoint Medical, Lunit Oncology, and Amazon Web Services (AWS) — for their generous support! ImFusion GmbH, ScreenPoint Medical, and Lunit Oncology made it possible for us to award the monetary awards to the top-performing teams 🏆

Special thanks goes to Amazon Web Services (AWS) for providing the compute resources that made this benchmark possible and Grand-Challenge.org for their continued technical support in hosting and running the competition, helping to make the UNICORN Challenge a true success.