Data
Loading…
Diagnosis results
Loading…
Test sets were never used to choose the model. Intervals are 95% bootstrap. Baselines: random (expected value), last model call, fallback order.
Queue
Running now
Machine
Model choice (grouped cross-validation on train + val)
Latest logs
Loading the dataset (about 10 s the first time)…
Choose filters and press Show.
Open a run from the Runs list.
Run
Diagnosis
…
Experiment
Steps
Orange edge: suspect named by the diagnosis model. Red edge: the known guilty step. "used" = earlier steps whose output this step read.
Fork at step
Replace this step's output and re-run. Steps before it come from the recording; later steps are re-used when unaffected.