Artifacts
What each model built on each task, shown on the held-out cases it was graded on.
Plume dispersion forecast
Each model forecast hourly ground-level concentrations at 24 receptors for 36-hour releases from a planned stack in a mountain valley, from a tracer campaign on site. These are two of the graded hidden test scenarios, each model's best and median attempt next to the true values; agents made no forecast on public data, so there is no public dev view.