Release snapshot

.gitattributes

@@ -0,0 +1,60 @@
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.avro filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.lz4 filter=lfs diff=lfs merge=lfs -text
+*.mds filter=lfs diff=lfs merge=lfs -text
+*.mlmodel filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.npy filter=lfs diff=lfs merge=lfs -text
+*.npz filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pickle filter=lfs diff=lfs merge=lfs -text
+*.pkl filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tar filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
+*.xz filter=lfs diff=lfs merge=lfs -text
+*.zip filter=lfs diff=lfs merge=lfs -text
+*.zst filter=lfs diff=lfs merge=lfs -text
+*tfevents* filter=lfs diff=lfs merge=lfs -text
+# Audio files - uncompressed
+*.pcm filter=lfs diff=lfs merge=lfs -text
+*.sam filter=lfs diff=lfs merge=lfs -text
+*.raw filter=lfs diff=lfs merge=lfs -text
+# Audio files - compressed
+*.aac filter=lfs diff=lfs merge=lfs -text
+*.flac filter=lfs diff=lfs merge=lfs -text
+*.mp3 filter=lfs diff=lfs merge=lfs -text
+*.ogg filter=lfs diff=lfs merge=lfs -text
+*.wav filter=lfs diff=lfs merge=lfs -text
+# Image files - uncompressed
+*.bmp filter=lfs diff=lfs merge=lfs -text
+*.gif filter=lfs diff=lfs merge=lfs -text
+*.png filter=lfs diff=lfs merge=lfs -text
+*.tiff filter=lfs diff=lfs merge=lfs -text
+# Image files - compressed
+*.jpg filter=lfs diff=lfs merge=lfs -text
+*.jpeg filter=lfs diff=lfs merge=lfs -text
+*.webp filter=lfs diff=lfs merge=lfs -text
+# Video files - compressed
+*.mp4 filter=lfs diff=lfs merge=lfs -text
+*.webm filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,78 @@
+---
+license: cc-by-sa-4.0
+pretty_name: Doc-Split Benchmark (page-stream segmentation)
+task_categories: [image-classification, text-classification]
+language: [multilingual]
+tags: [page-stream-segmentation, document-boundary-detection, document-ai, benchmark]
+size_categories: [1K<n<10K]
+configs:
+- config_name: our200
+  data_files: [{split: test, path: our200/test-*.parquet}]
+---
+
+# Doc-Split Benchmark
+
+The **evaluation slice** for page-stream segmentation — the exact set behind the
+[leaderboard](https://huggingface.co/spaces/nutrientdocs/doc-split-leaderboard) and the cloud-VLM
+comparison. Self-contained (page images embedded), with a reference scorer so results are reproducible.
+This is the *benchmark*, not the training corpus (which stays private).
+
+- 🏆 **Leaderboard:** [doc-split-leaderboard](https://huggingface.co/spaces/nutrientdocs/doc-split-leaderboard)
+- 🎯 **Demo:** [doc-split-demo](https://huggingface.co/spaces/nutrientdocs/doc-split-demo)
+- 🟢 **Model:** [doc-split-mini-e5](https://huggingface.co/nutrientdocs/doc-split-mini-e5) (open weights)
+- 🌍 **OpenPSS cuts:** [openpss-mirror](https://huggingface.co/datasets/nutrientdocs/openpss-mirror) (SHORT/LONG, self-contained)
+
+## Cuts
+
+| Cut | Where | Notes |
+|---|---|---|
+| **our-200** | this repo (`our200` config) | 200 license-clean synthetic-concat streams, ~78% boundary rate |
+| **OpenPSS short / long** | [openpss-mirror](https://huggingface.co/datasets/nutrientdocs/openpss-mirror) | Dutch; sparse- & dense-boundary regimes |
+| TABME++ / Tobacco800 | upstream (not redistributed) | reported for reference; licenses/gating prevent re-hosting |
+
+## Load
+
+```python
+from datasets import load_dataset
+ds = load_dataset("nutrientdocs/doc-split-benchmark", "our200", split="test")
+# one row = one page; group by stream_id, order by position.
+```
+
+## Schema
+
+| Field | Type | Meaning |
+|---|---|---|
+| `stream_id` | string | groups pages into one ordered stream |
+| `position` | int | page index within the stream (0-based) |
+| `boundary` | int | **1 = starts a new document** (position 0 always 1) |
+| `page_text` | string | page OCR text (may be empty) |
+| `image` | image | page image (512px) |
+| `source` | string | provenance |
+
+## How to score
+
+Metric: **boundary page-F1** (internal positions, page 0 forced) + Cohen's **κ**, WindowDiff, Pk.
+The reference scorer ships in this repo:
+
+```bash
+python score.py --pred your_preds.json      # preds: {stream_id: [0/1 per page]}
+```
+
+`score.py` uses `docsplit.eval.metrics.aggregate` (bundled). κ is reported because a degenerate
+predict-none/all can look fine on F1 but scores κ ≈ 0 on the sparse cuts. See `SUBMISSION.md` to add a row to
+the leaderboard.
+
+`manifest.json` records provenance (per-source stream/page counts, boundary rates, seed).
+
+## License & attribution
+
+CC-BY-SA-4.0 for the our-200 packaging (Nutrient-generated synthetic concatenations). OpenPSS cuts are in
+`openpss-mirror` under their upstream terms. TABME++/Tobacco800 are referenced, not redistributed.
+
+## About the author
+
+<a href="https://nutrient.io/">
+  <img src="https://avatars2.githubusercontent.com/u/1527679?v=3&s=200" height="80" />
+</a>
+
+This project is maintained and funded by [Nutrient](https://nutrient.io/) - The #1 PDF SDK library for viewing, editing, eSigning, and more.

SUBMISSION.md

@@ -0,0 +1,17 @@
+# 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`](https://huggingface.co/datasets/nutrientdocs/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](https://huggingface.co/spaces/nutrientdocs/doc-split-leaderboard)
+   adding `results/<model>-<cut>.json`:
+
+```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).

manifest.json

@@ -0,0 +1,16 @@
+{
+  "cut": "our200",
+  "streams": 200,
+  "pages": 894,
+  "boundaries": 702,
+  "boundary_rate": 0.7852,
+  "by_source": {
+    "openpss_short": 71,
+    "doclayout-render": 565,
+    "commonforms-synth": 170,
+    "cord-v2": 24,
+    "idl-wds": 22,
+    "pdfa-eng-wds": 14,
+    "sroie": 28
+  }
+}

our200/test-00000-of-00001.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e26fedbe482c3d4c7e21e9a0ce8f7848e5e1d1443722b63a62422e85d9be7010
+size 282433884

score.py

@@ -0,0 +1,90 @@
+"""Reference scorer for the Doc-Split Benchmark — self-contained (only needs `datasets`).
+
+Boundary page-F1 (internal positions, page 0 forced) + Cohen's kappa + WindowDiff + Pk.
+
+Usage:
+  python score.py --pred preds.json [--config our200]
+  preds.json = {stream_id: [0/1, ...]}  (one label per page, in position order; index 0 may be omitted/ignored)
+"""
+import argparse, json
+from collections import defaultdict
+
+
+def _counts(gold, pred):
+    tp = fp = fn = tn = 0
+    for g, p in zip(gold[1:], pred[1:]):            # skip page 0 (forced boundary for both)
+        if g and p: tp += 1
+        elif p and not g: fp += 1
+        elif g and not p: fn += 1
+        else: tn += 1
+    return tp, fp, fn, tn
+
+
+def _seg(b):
+    out, c = [], 0
+    for x in b:
+        c += 1 if x else 0; out.append(c)
+    return out
+
+
+def _k(ref):
+    n = len(ref); ns = max(1, sum(ref)); return max(1, round((n / ns) / 2))
+
+
+def windowdiff(g, p):
+    gg, pp = g[1:], p[1:]; m = len(gg)
+    if m == 0: return 0.0
+    k = min(_k(g), m); n = m - k + 1
+    if n <= 0: return float(sum(gg) != sum(pp))
+    return sum(1 for i in range(n) if sum(gg[i:i+k]) != sum(pp[i:i+k])) / n
+
+
+def pk(g, p):
+    n = len(g)
+    if n < 2: return 0.0
+    k = min(_k(g), n - 1); sg, sp = _seg(g), _seg(p); t = n - k
+    if t <= 0: return 0.0
+    return sum(1 for i in range(t) if (sg[i] == sg[i+k]) != (sp[i] == sp[i+k])) / t
+
+
+def aggregate(streams):
+    TP = FP = FN = TN = 0; em = wd = pkv = 0.0; n = len(streams)
+    for s in streams:
+        g, p = s["gold"], s["pred"]
+        tp, fp, fn, tn = _counts(g, p); TP += tp; FP += fp; FN += fn; TN += tn
+        em += 1.0 if list(g[1:]) == list(p[1:]) else 0.0
+        wd += windowdiff(g, p); pkv += pk(g, p)
+    prec = TP / (TP + FP) if TP + FP else 1.0
+    rec = TP / (TP + FN) if TP + FN else 1.0
+    f1 = 2 * prec * rec / (prec + rec) if prec + rec else 0.0
+    tot = TP + FP + FN + TN
+    po = (TP + TN) / tot if tot else 0.0
+    pe = ((TP + FP) / tot) * ((TP + FN) / tot) + ((FN + TN) / tot) * ((FP + TN) / tot) if tot else 0.0
+    kappa = (po - pe) / (1 - pe) if (1 - pe) > 1e-12 else 0.0
+    return {"f1": round(f1, 4), "precision": round(prec, 4), "recall": round(rec, 4),
+            "kappa": round(kappa, 4), "windowdiff": round(wd / n, 4), "pk": round(pkv / n, 4),
+            "exact_match": round(em / n, 4), "n_streams": n}
+
+
+def main():
+    ap = argparse.ArgumentParser()
+    ap.add_argument("--pred", required=True); ap.add_argument("--config", default="our200")
+    a = ap.parse_args()
+    from datasets import load_dataset
+    ds = load_dataset("nutrientdocs/doc-split-benchmark", a.config, split="test")
+    gold = defaultdict(list)
+    for r in ds:
+        gold[r["stream_id"]].append((r["position"], int(r["boundary"])))
+    preds = json.load(open(a.pred))
+    streams = []
+    for sid, pg in gold.items():
+        g = [b for _, b in sorted(pg)]
+        p = preds.get(sid)
+        if p is None:
+            p = [1] + [0] * (len(g) - 1)            # missing prediction -> forced page-0 only
+        streams.append({"gold": g, "pred": list(p)[:len(g)]})
+    print(json.dumps(aggregate(streams), indent=2))
+
+
+if __name__ == "__main__":
+    main()