Upload README.md

README.md

@@ -1,3 +1,149 @@
 ---
+dataset_info:
+  configs:
+    - config_name: default
+      data_files:
+        - split: train
+          path: data/all-scores.json
+      features:
+        - name: cancer
+          dtype: string
+        - name: model
+          dtype: string
+        - name: mean
+          dtype: float64
+        - name: sd
+          dtype: float64
+        - name: n
+          dtype: int64
+        - name: stability
+          dtype: string
+        - name: url
+          dtype: string
+        - name: runs
+          dtype: list
+      splits:
+        - name: train
+          num_bytes: 0
+          num_examples: 27
+  download_size: 0
+  dataset_size: 0
+task_categories:
+  - table-to-text
+  - question-answering
+  - text-classification
+language:
+  - en
+tags:
+  - benchmark
+  - clinical
+  - medical-ai
+  - llm-evaluation
+  - reproducible-research
+  - healthcare
 license: cc-by-nc-4.0
+pretty_name: Surgeon-Tested Clinical AI Benchmark (TH-CAB v1.1)
+size_categories:
+  - 1K<n<10K
 ---
+
+# Surgeon-Tested Clinical AI Benchmark (TH-CAB v1.1)
+
+An independent, **reproducible** evaluation of large language models (LLMs) on real, de-identified cancer cases — scored by a practicing surgeon item-by-item against current clinical guidelines.
+
+- **Homepage & full leaderboard:** https://tanhaosheng.asia
+- **Methodology (citable authority, TH-CAB v1.1):** https://tanhaosheng.asia/methodology/
+- **Open data layer:** https://tanhaosheng.asia/data/
+
+> This is a **benchmark / research dataset**, not clinical validation, and not medical advice. All patient data is de-identified. AI is an assistant to clinicians, not a replacement.
+
+## What makes this dataset different
+
+Most "AI beats doctor" headlines report a single run on a curated question. This dataset instead reports **20 repeated runs per model × case configuration**, hand-scored by a surgeon on a fixed **16-item / 100-point rubric** mapped to 2026 CBCS / CSCO 2024 / NCCN 2025 guidelines. The result is a *distribution* (mean ± SD), not a lucky single number — a clinical safety signal.
+
+## Dataset at a glance
+
+- **27 configuration rows** (model × cancer)
+- **574 scored runs** total
+- **13 model configurations** across families: DeepSeek (V4 Flash, V4 Pro, R1, Quick, Web Fast), GLM (5.2, 5V-Turbo, 4-Flash), Gemini 3.1 Flash Lite, Agnes-2.0-flash, GPT-5, Zhipu
+- **11 cancer types:** Breast, Cervical, Colorectal, Esophageal, Gastric, Liver, Lung, Ovarian, Pancreatic, Prostate, Thyroid
+
+## Schema
+
+Each row of `data/all-scores.json` has the following fields:
+
+| Field | Type | Description |
+|---|---|---|
+| `cancer` | string | Cancer type (one de-identified real case per type) |
+| `model` | string | Model name and interface (API / Web / agent wrapper) |
+| `mean` | float | Mean score on the 16-item / 100-point rubric (20-run mean) |
+| `sd` | float | Standard deviation across runs (null if not reported) |
+| `n` | int | Number of repeated runs (typically 20) |
+| `stability` | string | TH-CAB stability class: `Stable` / `Caution` / `Unstable` (null if not classified) |
+| `url` | string | Link to the benchmark article with raw runs |
+| `runs` | list | Placeholder for per-run score sheets (extended in linked articles) |
+
+## Methodology in one line (TH-CAB v1.1)
+
+1. One real, de-identified case per cancer type (chosen for guideline-relevant staging/treatment traps).
+2. One locked prompt template, identical for every model — no model-specific tuning.
+3. Fixed 16-item / 100-point rubric mapped to 2026 CBCS / CSCO 2024 / NCCN 2025.
+4. 20 repeated runs per configuration to measure stability, not a single lucky answer.
+5. Every run scored **by hand by a surgeon** (no LLM-as-judge, no auto-scoring).
+6. Conflicts disclosed in visible correction boxes, never silently merged.
+
+Cite the methodology page (https://tanhaosheng.asia/methodology/) when referencing any score.
+
+## Loading the data
+
+```python
+from datasets import load_dataset
+
+ds = load_dataset("YOUR_HF_USERNAME/surgeon-tested-clinical-ai-benchmark")
+print(ds["train"][0])
+# {'cancer': 'Breast', 'model': 'DeepSeek V4 Flash (API)', 'mean': 90.5, 'sd': 3.2, 'n': 20, 'stability': 'Stable', 'url': '/benchmark-repeats/deepseek-v4pro-vs-flash', 'runs': []}
+```
+
+Or simply read the JSON directly:
+
+```python
+import json
+rows = json.load(open("data/all-scores.json"))
+```
+
+## Example analysis
+
+```python
+import json
+rows = json.load(open("data/all-scores.json"))
+
+# Best-scoring configurations per cancer type
+best = {}
+for r in rows:
+    c = r["cancer"]
+    if c not in best or r["mean"] > best[c]["mean"]:
+        best[c] = r
+for c, r in best.items():
+    print(f"{c:12s} {r['model']:28s} {r['mean']}")
+```
+
+## License
+
+**CC-BY-NC 4.0** — attribution required, non-commercial use. Please cite the dataset and link back to https://tanhaosheng.asia.
+
+## Citation
+
+```bibtex
+@misc{tan2026surgeontested,
+  title        = {Surgeon-Tested Clinical AI Benchmark (TH-CAB v1.1)},
+  author       = {Tan, Haosheng},
+  year         = {2026},
+  publisher    = {Surgeon-Tested AI},
+  howpublished = {\url{https://tanhaosheng.asia}},
+  note         = {De-identified real cancer cases, 16-item/100 rubric, 20 repeated runs, surgeon-scored}
+}
+```
+
+## Disclaimer
+
+For educational and research purposes only, based on de-identified cases. **Not individual medical advice.** Consult a physician and follow current guidelines (NCCN / CSCO / CBCS). Nothing on this dataset or the associated site is a substitute for professional medical judgment.