add git address

DATASETS.md

@@ -0,0 +1,95 @@
+# Benchmark 데이터셋 배포 구성
+
+Hugging Face Dataset 저장소를 아래 단위로 관리합니다. 개인 계정에서 먼저 배포하고
+검증한 뒤, 필요하면 각 저장소를 `KETI-NLP` 조직으로 이전합니다.
+아래 저장소 ID와 경로는 배포할 때 사용할 제안이며, 아직 생성된 저장소를 의미하지 않습니다.
+
+| 표시 이름 | 개인 저장소 ID 제안 | 포함 데이터셋 |
+| --- | --- | --- |
+| 진실성 Benchmark | `DoolyKim22/Truthfulness-Benchmark` | 일반 상식, 수치 정보, 팩트 체크, 멀티 모달 |
+| 소버린 Benchmark | `DoolyKim22/Sovereign-Benchmark` | 소버린 |
+| K-Prism | `DoolyKim22/K-Prism` | K-Prism Text / Image 트랙 |
+
+진실성의 네 항목은 하나의 저장소 안에서 별도의 configuration으로 제공합니다.
+각 항목의 원본 필드와 정답 체계는 유지하며, 하나의 JSON으로 합치지 않습니다.
+이 구성은 데이터 배포 단위이며, 리더보드의 평가 항목이나 평균 계산은 변경하지 않습니다.
+
+## 진실성 Benchmark
+
+다음은 JSON 형식의 평가 데이터를 사용하는 폴더 구조 예시입니다.
+실제 파일명과 형식은 원본을 확인한 후 반영합니다.
+
+```text
+truthfulness-benchmark/
+├── README.md
+├── general-knowledge/
+│   └── test.json
+├── numerical/
+│   └── test.json
+├── fact-check/
+│   └── test.json
+└── multimodal/
+    ├── test.json
+    └── images/
+```
+
+해당 폴더의 README 상단에 아래 설정을 넣으면 네 항목을 따로 선택할 수 있습니다.
+경로는 위 예시에 해당하며, 실제 파일 경로와 일치시켜야 합니다.
+
+```yaml
+---
+language:
+- ko
+configs:
+- config_name: general_knowledge
+  data_files:
+  - split: test
+    path: general-knowledge/test.json
+- config_name: numerical
+  data_files:
+  - split: test
+    path: numerical/test.json
+- config_name: fact_check
+  data_files:
+  - split: test
+    path: fact-check/test.json
+- config_name: multimodal
+  data_files:
+  - split: test
+    path: multimodal/test.json
+---
+```
+
+## 소버린 Benchmark
+
+```text
+sovereign-benchmark/
+├── README.md
+└── test.json
+```
+
+원본이 여러 파일 또는 하위 평가 항목으로 나뉘어 있다면 그 구조를 유지하고,
+필요에 따라 소버린 저장소 안에도 configurations를 정의합니다.
+
+## 업로드
+
+1. <https://huggingface.co/new-dataset>에서 각 Dataset 저장소를 생성합니다.
+2. 공개할 원본 데이터와 데이터 설명 README를 각각의 배포 폴더에 준비합니다.
+   README에는 항목별 출처, 필드, 평가 방식, 라이선스를 기재합니다.
+3. 다음 명령의 `/실제/경로/` 부분을 준비한 폴더 경로로 바꿔 실행합니다.
+
+```bash
+hf upload DoolyKim22/Truthfulness-Benchmark \
+  "/실제/경로/truthfulness-benchmark" . --repo-type dataset
+
+hf upload DoolyKim22/Sovereign-Benchmark \
+  "/실제/경로/sovereign-benchmark" . --repo-type dataset
+```
+
+이미지는 annotation에서 참조하는 상대 경로를 유지하여 함께 업로드합니다.
+`static-space/front/data/benchmark.json`은 모델별 점수 파일이며 평가 원본이 아닙니다.
+현재 이 작업 폴더에서 다섯 항목의 원본 위치는 확인되지 않았습니다.
+원본 경로를 확인한 뒤 배포 폴더를 구성하고 이미지 참조와 데이터 형식을 검증해야 합니다.
+
+참고: [데이터셋 업로드](https://huggingface.co/docs/hub/datasets-adding),
+[여러 데이터셋 구성](https://huggingface.co/docs/hub/datasets-data-files-configuration).

README.md

@@ -6,88 +6,4 @@ colorTo: gray
 sdk: static
 app_file: index.html
 pinned: false
----
-
-# KETI Leaderboard
-
-A static leaderboard for published KETI model evaluations. The browser loads
-`front/data/benchmark.json` directly; no Python server, Docker runtime, access token,
-or build step is needed. Dataset selection, model search, provider filtering,
-comparison charts, theme selection, and CSV export are available.
-
-The initial snapshot contains the 19 models and the 9 displayed datasets retrieved
-from the existing public leaderboard during migration. Scores and evaluation
-timestamps are preserved. The snapshot records its source and capture time;
-internal evaluation job identifiers and error messages are not included.
-
-## Run locally
-
-From the repository root:
-
-```bash
-python3 -m http.server 7860 --bind 127.0.0.1
-```
-
-Open <http://127.0.0.1:7860/>. Use an HTTP server rather than opening `index.html`
-as a local file, because the browser loads JavaScript modules and JSON.
-Tailwind, Chart.js, and fonts currently load from their existing public CDNs.
-
-## Verify the static page
-
-With Playwright and Chrome/Chromium installed, run:
-
-```bash
-python3 -m unittest discover -s tests -v
-```
-
-Set `CHROME_PATH` if the browser is not on `PATH`. These checks use a temporary
-local HTTP server and cover published scores and means, filters, comparison charts,
-CSV export, mobile controls, preferences, loading failures, and malformed data.
-Internet access is needed for the existing CDN assets in the chart test.
-
-## Update published results
-
-1. Evaluate models outside this Space.
-2. Update `front/data/benchmark.json` with the resulting scores.
-3. Commit and push the changed data file to the Space. Visitors can use **Refresh**
-   to load the published version again.
-
-The data file accepts either the legacy `benchmark.json` array of models or an
-object with `models` and an optional `datasets` array. Each model has `provider`,
-`name`, `repo`, `is_multimodal`, `updated_at`, and `scores`. Each score has
-`dataset_name`, numeric `score`, and `metric_type` (`raw` or `llm-as-judge`).
-The `datasets` array sets the displayed columns and their order; without it,
-the canonical datasets and any additional scored datasets are shown.
-
-The mean calculation retains the previous behavior: all selected, supported
-datasets must have scores. Unsupported multimodal datasets are excluded for
-text-only models. Missing means are displayed as `—`. A failed data request shows
-an error; it never substitutes generated scores. A failed refresh keeps the last
-successfully loaded results on screen.
-
-For an independently updated, **public** JSON source, change `dataUrl` in
-`front/config.mjs` to its HTTPS URL. That server must allow browser cross-origin
-requests (CORS). No automatic fallback to another source is applied. The default
-bundled snapshot remains independent of the personal Hugging Face account.
-All browser configuration and bundled files are public: do not put tokens or
-private evaluation data in them.
-
-## Deploy as a new Hugging Face Space
-
-This directory is an independent Git repository containing only the static app.
-The existing Docker Space at
-https://huggingface.co/spaces/DoolyKim22/KETI_Leaderboard remains separate.
-
-The deployment target is the new organization-owned Space
-[KETI-NLP/Benchmark_Leaderboard](https://huggingface.co/spaces/KETI-NLP/Benchmark_Leaderboard),
-using the **Static** SDK. The README metadata already specifies
-`sdk: static` and `app_file: index.html` as described in the
-[Static Spaces documentation](https://huggingface.co/docs/hub/spaces-sdks-static).
-Do not use the existing Docker Space as this repository's remote.
-
-The snapshot is independent of the running Docker app. Updates published to the
-old app after the snapshot was captured are not automatically copied here.
-Update `front/data/benchmark.json` and publish a new commit to refresh this app.
-
-Model submissions, automatic evaluations, and dataset backup workers are not part
-of the static site. Browser configuration lives in `front/config.mjs`.
+---

index.html

@@ -129,9 +129,10 @@
   <!-- Main -->
   <main class="mx-auto max-w-7xl px-4 md:px-6 py-6 md:py-8 space-y-6 md:space-y-8">
     <!-- Tabs -->
-    <div class="flex gap-2 md:gap-3">
+    <div class="flex flex-wrap gap-2 md:gap-3">
       <button class="tab tab-active" data-tab="leaderboard">🏆 Leaderboard</button>
       <button class="tab" data-tab="about">ℹ️ About</button>
+      <a class="btn btn-ghost" href="https://github.com/alsgur0720/K-Prism" target="_blank" rel="noopener noreferrer">↗ K-Prism Code</a>
     </div>
 
     <div id="dataStatus" class="card p-4 text-sm" role="status" aria-live="polite" data-state="loading">Loading published results…</div>
@@ -230,6 +231,10 @@
         <p class="text-sm text-slate-600 dark:text-slate-300">
           This leaderboard presents published model evaluations on KETI's ethicality and veracity benchmark datasets.
         </p>
+        <p class="text-sm text-slate-600 dark:text-slate-300">
+          Detailed K-Prism evaluation code and documentation:
+          <a class="underline" href="https://github.com/alsgur0720/K-Prism" target="_blank" rel="noopener noreferrer">alsgur0720/K-Prism on GitHub</a>.
+        </p>
         <p class="text-sm text-slate-600 dark:text-slate-300">
           Model submissions and automatic evaluations are not available on this page. Results are updated when the maintainers publish a new evaluation.
         </p>

upload_space.py

@@ -0,0 +1,53 @@
+"""Upload this static app to an existing Static Space using the saved HF login."""
+
+import argparse
+from pathlib import Path
+
+from huggingface_hub import HfApi
+
+
+ROOT = Path(__file__).resolve().parent
+FILES = [
+    '.gitignore',
+    'README.md',
+    'index.html',
+    'front/config.mjs',
+    'front/data.mjs',
+    'front/leaderboard.mjs',
+    'front/data/benchmark.json',
+    'tests/test_static_space.py',
+    'upload_space.py',
+]
+
+
+def main():
+    parser = argparse.ArgumentParser(description=__doc__)
+    parser.add_argument('--repo-id', default='DoolyKim22/Benchmark_Leaderboard')
+    parser.add_argument('--message', default='Update static benchmark leaderboard')
+    args = parser.parse_args()
+
+    for name in FILES:
+        if not (ROOT / name).is_file():
+            parser.error('Missing deployment file: ' + name)
+
+    api = HfApi()
+    space = api.space_info(args.repo_id)
+    if space.sdk != 'static':
+        parser.error('Upload target must already be a Static Space: ' + args.repo_id)
+
+    # upload_folder writes to the existing repository without create_repo().
+    # Older hf upload versions send space_sdk="gradio" in that extra request.
+    commit = api.upload_folder(
+        repo_id=args.repo_id,
+        repo_type='space',
+        folder_path=ROOT,
+        allow_patterns=FILES,
+        parent_commit=space.sha,
+        commit_message=args.message,
+    )
+    print('Uploaded:', commit.commit_url)
+    print('Space: https://huggingface.co/spaces/' + args.repo_id)
+
+
+if __name__ == '__main__':
+    main()