Explore the visual operations behind Omni-Reason. Inspect dataset inputs, targets, and prompts across all 340 task types.
Input and Target are dataset examples, not model predictions. Select a task to explore its prompt and benchmark.