SUBMISSION.md

Submit to the Doc-Split leaderboard

1. Predict boundaries for every stream in each cut: {stream_id: [0/1 per page, in position order]} (1 = page starts a new document; page 0 is a forced boundary). Cuts: our200 (this repo) and the OpenPSS SHORT/LONG cuts from openpss-mirror.
2. Score per cut: python score.py --pred preds.json --config our200 → {f1, kappa, windowdiff, pk, ...}.
3. Open a PR on the leaderboard Space adding results/<model>-<cut>.json:

{ "name": "your-model", "model": "org/your-model", "cut": "openpss_short",
  "open": true, "n": 108, "f1": 0.00, "kappa": 0.00,
  "hardware": "A100", "inputs": "image+text" }

Report κ alongside F1 — on the sparse cuts a degenerate predict-none/all scores κ ≈ 0 despite a misleading F1. Note whether your model saw these domains in training (in-domain vs zero-shot).