Task overview¶

The UNICORN Challenge currently includes one vision-language task, with additional tasks actively under development and expected to be added in the future.

In the existing task, participants are asked to generate a descriptive caption from a whole-slide pathology image using a vision-language model. This task is designed to test the model's ability to interpret visual data and summarize clinical insights, enabling potential applications in automated diagnostics and pathology report generation. The model should be able to understand features of the slide, such as tissue type, cellular morphology, and any prominent pathological findings, and provide a clear and concise textual description.

Participants will submit a Docker container, containing a pre-trained model ready to process and caption whole-slide images. Detailed guidelines on Docker setup and submission will be provided in the future.