Publish Ropedia Xperience-10M derived artifacts

EVIDENCE_CONTRACT.md

@@ -13,6 +13,7 @@ local artifact that a reader can inspect before trusting the dashboard.
 | Four extra direction probes are coded and evaluated. | `results/episode_task_suite/research_direction_extensions/research_direction_extension_results.json`, `docs/data/research_direction_extensions.json` | Verified single-episode probes | Not full human modeling, neural rendering, intent modeling, or world modeling solutions |
 | Qwen3-Omni infrastructure has passed technical smoke checks. | `results/omni_finetune/RUN_REPORT.md`, `results/omni_finetune/dataset_manifest.json`, `results/omni_finetune/metrics_eval.json` | Smoke-only evidence | One episode, 128 train windows; not a 32-episode pilot |
 | The real 32-episode LoRA pilot is blocked on gated data access, not on repo presentation. | `results/omni_finetune/DATA_BLOCKER_REPORT.md`, `results/omni_finetune/A100_HF_RELAY_STATUS.md`, `results/omni_finetune/source_discovery.json` | Blocker documented | No 32-episode metric should be claimed until the gate passes |
+| The public GitHub and Hugging Face bundles are publication-clean. | `scripts/validate_publication_package.py`, `docs/data/publication_audit.json` | Verified pass | Checks public files and HF bundles, not arbitrary ignored local scratch outputs |
 
 ## Review Order
 
@@ -25,4 +26,5 @@ local artifact that a reader can inspect before trusting the dashboard.
    neural heads under the same splits.
 5. Inspect `results/omni_finetune/DATA_BLOCKER_REPORT.md` before interpreting
    any Qwen3-Omni artifact.
-
+6. Inspect `docs/data/publication_audit.json` before publishing or sharing the
+   project externally.

PROJECT_README.md

@@ -40,10 +40,13 @@ This repo is organized around an explicit proof boundary:
 | Neural heads | `scripts/neural_task_models.py`, `results/episode_task_suite/neural_mlp/` | compact MLP heads, not a foundation model |
 | Research directions | `research_direction_taxonomy.json`, extension probe results | direct/proxy/diagnostic evidence, not full solutions |
 | Qwen3-Omni | `results/omni_finetune/DATA_BLOCKER_REPORT.md`, `A100_HF_RELAY_STATUS.md` | smoke-only until 32 valid episodes are available |
+| Publication hygiene | `scripts/validate_publication_package.py`, `docs/data/publication_audit.json` | public repo and HF bundles only; ignored local scratch files are excluded |
 
 Read the full contract in [`EVIDENCE_CONTRACT.md`](EVIDENCE_CONTRACT.md), or
 consume the machine-readable copy at
 [`docs/data/evidence_contract.json`](docs/data/evidence_contract.json).
+The current publication audit is at
+[`docs/data/publication_audit.json`](docs/data/publication_audit.json).
 
 ## Dataset Modality Coverage
 
@@ -75,6 +78,7 @@ Hugging Face Space app:
 | Minimal heads | softmax, ridge projection/regression, multi-label logistic heads | Keeps every input/output contract visible and debuggable |
 | Neural heads | PyTorch MLP classifiers/regressors under `neural_mlp/` | Checks whether nonlinear heads improve each task without changing features |
 | Evidence | metrics, predictions, confusion matrices, diagrams, dashboard | Makes the single-episode claims reviewable without rerunning first |
+| Publication audit | `docs/data/publication_audit.json` | Confirms public bundles contain no raw Xperience-10M data, Python caches, heavy archives, or token strings |
 
 ## Links
 
@@ -135,6 +139,7 @@ scripts/
   generate_visualizations.py        # refreshes SVG charts + summary JSON
   render_task_suite_infographic.py  # renders the ChatGPT-image-backed PNG
   render_overview_figures.py        # renders polished pipeline/architecture PNGs
+  validate_publication_package.py   # checks public repo + HF bundle hygiene
   omni/
     download_sample_modelscope.py   # mainland-China friendly sample download
     build_episode_manifest.py       # metadata-only multi-episode scanner
@@ -157,6 +162,7 @@ docs/
   index.html                        # GitHub Pages dashboard
   data/summary_metrics.json         # website-readable metrics bundle
   data/evidence_contract.json       # machine-readable proof boundary
+  data/publication_audit.json       # machine-readable publication hygiene check
   data/research_directions.json     # four-track website data bundle
   data/research_direction_extensions.json # four extra probe data bundle
   data/task_walkthroughs.json       # beginner task explanation data bundle

README.md

@@ -54,17 +54,18 @@ This is the reviewable half of the project. You can inspect the task outputs, co
 | Neural heads | `results/episode_task_suite/neural_mlp/` | compact MLP heads, not a foundation model |
 | Research directions | `research_direction_taxonomy.json`, extension probe results | direct/proxy/diagnostic evidence, not full solutions |
 | Qwen3-Omni | `DATA_BLOCKER_REPORT.md`, `A100_HF_RELAY_STATUS.md` | smoke-only until 32 valid episodes are available |
+| Publication hygiene | `docs/data/publication_audit.json`, `scripts/validate_publication_package.py` | public files and HF bundles only |
 
 ## What Is Included
 
 - `results/**/*.json`: verified metrics and metadata for minimal and neural MLP runs
 - `results/**/*.csv`: predictions, confusion matrices, per-class metrics, windows, boundaries
-- `results/**/*.npz`: compact derived neural prediction arrays for ranking/regression tasks
 - `results/**/history.json`: neural MLP training traces
 - `docs/assets/*.svg` and `docs/assets/*.png`: generated diagrams, charts, and ChatGPT-image-backed overview figures
 - `docs/assets/task_suite_infographic.png`: ChatGPT-image-backed infographic with larger public-sample modality atlas thumbnails, including audio waveform context, and verified metric overlays
 - `docs/data/summary_metrics.json`: dashboard-readable summary bundle
 - `docs/data/evidence_contract.json`: machine-readable proof boundary
+- `docs/data/publication_audit.json`: machine-readable publication hygiene check
 - `docs/data/research_directions.json`: generated four-track taxonomy for the website
 - `docs/data/research_direction_extensions.json`: four extra data-backed probes, one per research direction
 - `docs/data/task_walkthroughs.json`: beginner-oriented input/process/output guide for all 12 tasks
@@ -72,9 +73,13 @@ This is the reviewable half of the project. You can inspect the task outputs, co
 - `results/episode_task_suite/research_direction_extensions/`: metrics, prediction CSVs, rank CSVs, and Markdown summary for the four extension probes
 - `results/episode_task_suite/task_walkthroughs/`: case-study walkthroughs for every task contract
 - `scripts/*.py`: reproduction scripts
+- `scripts/validate_publication_package.py`: public bundle validator
 - `notes/*.md`: interpretation and reproducibility notes
 
-The companion model repo stores the lightweight model checkpoints and mirrors the binary arrays:
+The companion model repo stores the lightweight model checkpoints and mirrors
+the binary arrays (`model.npz`, `model.pt`, and compact neural prediction
+arrays). This artifact dataset stays focused on reviewable CSV/JSON/Markdown,
+scripts, notes, and visual assets:
 
 https://huggingface.co/cy0307/ropedia-xperience-10m-task-baselines
 

docs/data/publication_audit.json

@@ -0,0 +1,97 @@
+{
+  "status": "pass",
+  "generated_at_utc": "2026-05-31T22:03:21+00:00",
+  "checks": [
+    {
+      "name": "required_publication_assets_present",
+      "status": "pass",
+      "missing": []
+    },
+    {
+      "name": "no_generated_python_caches",
+      "status": "pass",
+      "count": 0
+    },
+    {
+      "name": "no_raw_xperience10m_data",
+      "status": "pass",
+      "count": 0
+    },
+    {
+      "name": "no_heavy_model_archives",
+      "status": "pass",
+      "count": 0
+    },
+    {
+      "name": "no_hf_tokens_in_public_text",
+      "status": "pass",
+      "count": 0
+    }
+  ],
+  "required_assets": {
+    "README.md": true,
+    "EVIDENCE_CONTRACT.md": true,
+    "DATA_NOTICE.md": true,
+    "docs/index.html": true,
+    "docs/data/evidence_contract.json": true,
+    "docs/data/summary_metrics.json": true,
+    "docs/assets/task_suite_infographic.png": true,
+    "docs/assets/pipeline_diagram.png": true,
+    "docs/assets/task_architectures.png": true,
+    "results/episode_task_suite/summary_report.json": true,
+    "results/episode_task_suite/feature_manifest.json": true,
+    "results/episode_task_suite/neural_mlp/timeline_action/metrics.json": true,
+    "results/omni_finetune/DATA_BLOCKER_REPORT.md": true,
+    "results/omni_finetune/A100_HF_RELAY_STATUS.md": true,
+    "scripts/episode_task_suite.py": true,
+    "scripts/neural_task_models.py": true,
+    "scripts/omni/train_qwen3_omni_lora.py": true
+  },
+  "scans": {
+    "github_repo": {
+      "root": "/Users/chaoyue/Documents/Codex/2026-05-29/i-am-learning-this-dataset-https/working_repo_copy",
+      "exists": true,
+      "file_count": 257,
+      "text_file_count": 197,
+      "largest_file": {
+        "path": "results/episode_task_suite/modality_reconstruction/predictions.npz",
+        "bytes": 52601010
+      },
+      "violations": []
+    },
+    "hf_space_bundle": {
+      "root": "/Users/chaoyue/Documents/Codex/2026-05-29/i-am-learning-this-dataset-https/hf_publish/space",
+      "exists": true,
+      "file_count": 25,
+      "text_file_count": 8,
+      "largest_file": {
+        "path": "assets/pipeline_diagram_base.png",
+        "bytes": 2279984
+      },
+      "violations": []
+    },
+    "hf_artifact_bundle": {
+      "root": "/Users/chaoyue/Documents/Codex/2026-05-29/i-am-learning-this-dataset-https/hf_publish/artifacts",
+      "exists": true,
+      "file_count": 220,
+      "text_file_count": 171,
+      "largest_file": {
+        "path": "results/episode_task_suite/neural_mlp/temporal_order/model.pt",
+        "bytes": 13406129
+      },
+      "violations": []
+    },
+    "hf_model_bundle": {
+      "root": "/Users/chaoyue/Documents/Codex/2026-05-29/i-am-learning-this-dataset-https/hf_publish/model",
+      "exists": true,
+      "file_count": 162,
+      "text_file_count": 116,
+      "largest_file": {
+        "path": "artifacts/episode_task_suite/cross_modal_retrieval/model.npz",
+        "bytes": 41310574
+      },
+      "violations": []
+    }
+  },
+  "violations": []
+}

docs/index.html

@@ -370,6 +370,7 @@
     .evidence-card h3 { margin: 0; font-size: 18px; line-height: 1.2; }
     .evidence-card p { margin: 0; color: var(--muted); line-height: 1.6; }
     .evidence-card code { color: var(--ink); font-family: var(--font-mono); font-size: 12px; }
+    .evidence-card:last-child { grid-column: 1 / -1; }
     .evidence-links {
       display: flex;
       flex-wrap: wrap;
@@ -632,6 +633,7 @@
     @media (max-width: 640px) {
       .wrap { width: min(100% - 28px, var(--max)); }
       .hero-stats, .models, .task-grid, .artifact-grid, .evidence-grid, .chart-grid, .callout-row, .direction-grid, .baseline-strip, .extension-grid, .walkthrough-grid, .walk-flow { grid-template-columns: 1fr; }
+      .evidence-card:last-child { grid-column: auto; }
       .hero-inner, section { padding: 46px 0; }
       .signal { grid-template-columns: 1fr; }
       .signal strong { text-align: left; }
@@ -759,6 +761,15 @@
               <a href="data/evidence_contract.json">machine JSON</a>
             </div>
           </article>
+          <article class="evidence-card">
+            <span class="status-pill">verified</span>
+            <h3>Publication bundles pass hygiene checks</h3>
+            <p>The validator checks required assets, raw-data exclusion, Python cache exclusion, heavy archive exclusion, and accidental HF token strings across GitHub and the HF bundles.</p>
+            <div class="evidence-links">
+              <a href="data/publication_audit.json">publication audit</a>
+              <a href="https://github.com/ChaoYue0307/ropedia-xperience-10m-task-suite/blob/main/scripts/validate_publication_package.py">validator script</a>
+            </div>
+          </article>
         </div>
       </div>
     </section>

scripts/validate_publication_package.py

@@ -0,0 +1,225 @@
+#!/usr/bin/env python3
+"""Validate public publication hygiene for the repo and HF bundles.
+
+This check is intentionally conservative: it scans the GitHub repo plus the
+prepared Hugging Face Space/artifact/model folders for generated Python caches,
+raw Xperience-10M data, heavyweight checkpoint formats that should not be
+published here, and accidental Hugging Face token strings.
+"""
+
+from __future__ import annotations
+
+import argparse
+import json
+import re
+import subprocess
+import sys
+from datetime import datetime, timezone
+from pathlib import Path
+
+
+ROOT = Path(__file__).resolve().parents[1]
+DEFAULT_HF_ROOT = ROOT.parent / "hf_publish"
+
+BANNED_DIR_NAMES = {"__pycache__"}
+BANNED_FILE_NAMES = {".DS_Store"}
+BANNED_SUFFIXES = {".pyc", ".pyo"}
+RAW_DATA_SUFFIXES = {".mp4", ".hdf5", ".h5", ".rrd"}
+HEAVY_MODEL_SUFFIXES = {".safetensors", ".bin", ".tar"}
+TEXT_SUFFIXES = {
+    ".csv",
+    ".html",
+    ".json",
+    ".md",
+    ".py",
+    ".sh",
+    ".txt",
+    ".yaml",
+    ".yml",
+}
+TOKEN_PATTERN = re.compile(r"hf_[A-Za-z0-9]{20,}")
+
+
+def rel(path: Path, base: Path) -> str:
+    try:
+        return path.relative_to(base).as_posix()
+    except ValueError:
+        return path.as_posix()
+
+
+def git_public_paths(root: Path) -> list[Path] | None:
+    try:
+        result = subprocess.run(
+            ["git", "-C", str(root), "ls-files", "--cached", "--others", "--exclude-standard"],
+            check=True,
+            stdout=subprocess.PIPE,
+            stderr=subprocess.DEVNULL,
+            text=True,
+        )
+    except (OSError, subprocess.CalledProcessError):
+        return None
+    return [root / line for line in result.stdout.splitlines() if line.strip()]
+
+
+def iter_public_files(root: Path, paths: list[Path] | None = None):
+    if paths is not None:
+        for path in paths:
+            if path.exists():
+                yield path
+        return
+    if not root.exists():
+        return
+    for path in root.rglob("*"):
+        parts = set(path.parts)
+        if ".git" in parts or ".venv" in parts or "venv" in parts:
+            continue
+        yield path
+
+
+def scan(root: Path, *, paths: list[Path] | None = None) -> dict:
+    violations: list[dict] = []
+    text_files = 0
+    total_files = 0
+    largest_file = {"path": None, "bytes": 0}
+
+    for path in iter_public_files(root, paths):
+        path_rel = rel(path, root)
+        if path.is_dir():
+            if path.name in BANNED_DIR_NAMES:
+                violations.append({"kind": "generated_cache_dir", "path": path_rel})
+            continue
+
+        total_files += 1
+        size = path.stat().st_size
+        if size > largest_file["bytes"]:
+            largest_file = {"path": path_rel, "bytes": size}
+
+        suffix = path.suffix.lower()
+        if path.name in BANNED_FILE_NAMES or suffix in BANNED_SUFFIXES:
+            violations.append({"kind": "generated_cache_file", "path": path_rel})
+        if suffix in RAW_DATA_SUFFIXES:
+            violations.append({"kind": "raw_xperience10m_data", "path": path_rel})
+        if suffix in HEAVY_MODEL_SUFFIXES:
+            violations.append({"kind": "heavy_model_or_archive", "path": path_rel})
+
+        if suffix in TEXT_SUFFIXES:
+            text_files += 1
+            try:
+                text = path.read_text(encoding="utf-8", errors="ignore")
+            except OSError:
+                continue
+            if TOKEN_PATTERN.search(text):
+                violations.append({"kind": "possible_hf_token", "path": path_rel})
+
+    return {
+        "root": str(root),
+        "exists": root.exists(),
+        "file_count": total_files,
+        "text_file_count": text_files,
+        "largest_file": largest_file,
+        "violations": violations,
+    }
+
+
+def required_assets(root: Path) -> dict[str, bool]:
+    required = [
+        "README.md",
+        "EVIDENCE_CONTRACT.md",
+        "DATA_NOTICE.md",
+        "docs/index.html",
+        "docs/data/evidence_contract.json",
+        "docs/data/summary_metrics.json",
+        "docs/assets/task_suite_infographic.png",
+        "docs/assets/pipeline_diagram.png",
+        "docs/assets/task_architectures.png",
+        "results/episode_task_suite/summary_report.json",
+        "results/episode_task_suite/feature_manifest.json",
+        "results/episode_task_suite/neural_mlp/timeline_action/metrics.json",
+        "results/omni_finetune/DATA_BLOCKER_REPORT.md",
+        "results/omni_finetune/A100_HF_RELAY_STATUS.md",
+        "scripts/episode_task_suite.py",
+        "scripts/neural_task_models.py",
+        "scripts/omni/train_qwen3_omni_lora.py",
+    ]
+    return {item: (root / item).exists() for item in required}
+
+
+def build_report(hf_root: Path) -> dict:
+    roots = {
+        "github_repo": ROOT,
+        "hf_space_bundle": hf_root / "space",
+        "hf_artifact_bundle": hf_root / "artifacts",
+        "hf_model_bundle": hf_root / "model",
+    }
+    scans = {}
+    for name, path in roots.items():
+        public_paths = git_public_paths(path) if name == "github_repo" else None
+        scans[name] = scan(path, paths=public_paths)
+    assets = required_assets(ROOT)
+    missing_assets = [path for path, present in assets.items() if not present]
+    violations = [
+        {"root": name, **violation}
+        for name, result in scans.items()
+        for violation in result["violations"]
+    ]
+    checks = [
+        {
+            "name": "required_publication_assets_present",
+            "status": "pass" if not missing_assets else "fail",
+            "missing": missing_assets,
+        },
+        {
+            "name": "no_generated_python_caches",
+            "status": "pass"
+            if not any(v["kind"].startswith("generated_cache") for v in violations)
+            else "fail",
+            "count": sum(1 for v in violations if v["kind"].startswith("generated_cache")),
+        },
+        {
+            "name": "no_raw_xperience10m_data",
+            "status": "pass" if not any(v["kind"] == "raw_xperience10m_data" for v in violations) else "fail",
+            "count": sum(1 for v in violations if v["kind"] == "raw_xperience10m_data"),
+        },
+        {
+            "name": "no_heavy_model_archives",
+            "status": "pass" if not any(v["kind"] == "heavy_model_or_archive" for v in violations) else "fail",
+            "count": sum(1 for v in violations if v["kind"] == "heavy_model_or_archive"),
+        },
+        {
+            "name": "no_hf_tokens_in_public_text",
+            "status": "pass" if not any(v["kind"] == "possible_hf_token" for v in violations) else "fail",
+            "count": sum(1 for v in violations if v["kind"] == "possible_hf_token"),
+        },
+    ]
+    status = "pass" if all(check["status"] == "pass" for check in checks) else "fail"
+    return {
+        "status": status,
+        "generated_at_utc": datetime.now(timezone.utc).isoformat(timespec="seconds"),
+        "checks": checks,
+        "required_assets": assets,
+        "scans": scans,
+        "violations": violations,
+    }
+
+
+def main() -> int:
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--hf-root", type=Path, default=DEFAULT_HF_ROOT)
+    parser.add_argument("--output", type=Path, default=ROOT / "docs/data/publication_audit.json")
+    args = parser.parse_args()
+
+    report = build_report(args.hf_root.resolve())
+    args.output.parent.mkdir(parents=True, exist_ok=True)
+    args.output.write_text(json.dumps(report, indent=2) + "\n", encoding="utf-8")
+    print(f"{report['status'].upper()}: wrote {args.output}")
+    if report["status"] != "pass":
+        for violation in report["violations"][:40]:
+            print(f"- {violation['root']}: {violation['kind']} {violation['path']}")
+        if len(report["violations"]) > 40:
+            print(f"- ... {len(report['violations']) - 40} more violations")
+        return 1
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())