Upload README.md with huggingface_hub

README.md

@@ -27,88 +27,6 @@ The public subset contains 120 tasks across Global (English) and Chinese markets
 - 💻 Code: <https://github.com/ICBCBench/ICBCBench>
 - 📊 Dataset: <https://huggingface.co/datasets/DeepFin-Intelligence/ICBCBench>
 
-## Repository Layout
-
-```
-├── app.py                          # Gradio entry point
-├── create_leaderboard.py           # Main Gradio Blocks UI
-├── tabs/
-│   ├── leaderboard_tab.py          # Leaderboard rendering
-│   ├── data_viewer_tab.py          # Single-model report viewer
-│   ├── data_viewer_side_by_side_tab.py  # Side-by-side comparison
-│   └── shared_data.py              # Lazy data loading
-├── utils/
-│   ├── rank_leaderboard.py         # Aggregate raw results -> leaderboard.csv
-│   └── merge_raw_data.py           # Merge raw data + scores -> data_viewer.jsonl
-└── data/
-    ├── leaderboard.csv             # Aggregated leaderboard (generated)
-    ├── data_viewer.jsonl           # Per-sample viewer data (generated)
-    ├── raw_data/                   # Per-model prompts + articles
-    └── raw_results/                # Per-model evaluation scores
-```
-
-## Data Format
-
-### `data/leaderboard.csv`
-
-| Column | Description |
-|--------|-------------|
-| `model` | Model / system name |
-| `overall` | Weighted overall score across EN & ZH, Objective & Subjective |
-| `objective_en` | Objective score on Global (English) tasks |
-| `objective_zh` | Objective score on Chinese tasks |
-| `objective_avg` | Average objective score across both language tracks |
-| `subjective_en` | Subjective report score on Global (English) tasks |
-| `subjective_zh` | Subjective report score on Chinese tasks |
-| `subjective_avg` | Average subjective score across both language tracks |
-| `citation_score` | Citation consistency score (0-100) |
-| `source_quality` | Source authority & timeliness score (0-100) |
-| `rmsce` | Root Mean Square Calibration Error for objective confidence (lower is better) |
-
-### `data/raw_results/<model>/results.jsonl`
-
-Each line should be a JSON object:
-
-```json
-{
-  "id": "1",
-  "language": "en",
-  "track": "objective",
-  "score": 0.85,
-  "correct": true,
-  "confidence": 0.90
-}
-```
-
-or for subjective tasks:
-
-```json
-{
-  "id": "41",
-  "language": "zh",
-  "track": "subjective",
-  "score": 0.72,
-  "expert_score": 0.74,
-  "citation_score": 0.68,
-  "source_quality_score": 0.70
-}
-```
-
-Scores can be either 0-1 ratios or 0-100 percentages.
-
-## Local Development
-
-```bash
-pip install -r requirements.txt
-python utils/rank_leaderboard.py      # generate data/leaderboard.csv
-python utils/merge_raw_data.py        # generate data/data_viewer.jsonl
-python app.py
-```
-
-> **Note**: The `data/raw_results/` directory in this repository currently contains legacy DeepResearch-Bench-format `race_result.txt` files. Before running `rank_leaderboard.py` for ICBCBench, replace these with ICBCBench `results.jsonl` files as described above. The checked-in `data/leaderboard.csv` is a sample ICBCBench leaderboard and will be overwritten if you rerun the script without updating `raw_results/`.
-
-Then open <http://localhost:7860>.
-
 ## Citation
 
 ```bibtex