Pick a dataset
Each dataset has its own dashboard — runs, progress curves, and comparison figures load only for the dataset you choose, so the page stays fast.
loading datasets…
Dataset
Best headroom per run vs iteration (running max).
1.0 = matched the oracle on the in-loop val subset.
loading…
Runs
Each card is one autoresearch run (<solver>-YYYYMMDD-NN)
with its best-iter validation and held-out
test figures. Click to expand the per-iteration timeline.
loading runs…
Run
Progress curve
Iterations
Stage checks (every 30 iter)
Shared scratch pad
Cross-run observations for this dataset (most recent first, capped).