Complete source-based prediction comparison
Running. This frozen study compares complete prediction ranges across all 48 original model, processing and mounting regions. It reuses generated data; it does not refit the 53 aircraft stretches.
The source/domain and joint-adapter checks have passed. The study keeps the original physical bounds, training observations and prediction-width targets. It uses four single-thread workers and is expected to take about 65–80 minutes, with a fixed two-hour ceiling. Results will replace this page automatically after completion, including failures or an unmet target.