Update README.md

README.md

@@ -1,5 +1,49 @@
 ---
+language:
+- en
 license: mit
+task_categories:
+- text-generation
+- text-classification
+tags:
+- education
+- cefr
+- synthetic-data
+- linguistics
+pretty_name: English CEFR Explorer Benchmark
+size_categories:
+- n<1K
+configs:
+- config_name: default
+  data_files:
+  - split: train
+    path: final_english_dataset.jsonl
 ---
-datasets:
-  - yasincicek/English-CEFR-Instruction-Tuning-Benchmark
+
+# 🇬🇧 English CEFR Explorer Benchmark
+
+This dataset is an automatically updated benchmark for testing LLM adherence to CEFR (Common European Framework of Reference for Languages) constraints.
+
+## Dataset Structure
+
+The dataset is formatted as a JSONL file optimized for **Instruction Tuning**.
+
+### Data Fields
+
+- **`id`**: Unique identifier for the generation task.
+- **`messages`**: Standard chat format (System, User, Assistant).
+  - `System`: Defines the persona (ESL Teacher).
+  - `User`: The constraint-based prompt (e.g., "Write B1 text using 'pollution'").
+  - `Assistant`: The generated output text.
+- **`scientific_metadata`**: (or `meta`) specific linguistic constraints used for validation.
+  - `target_level`: The intended CEFR level (A1-C2).
+  - `topic`: The semantic domain.
+  - `constraints`: The specific vocabulary/grammar rules enforced.
+
+## Automation
+
+This dataset is **Self-Expanding**. 
+Every day at 06:00 UTC, a GitHub Action generates new samples using Google Gemini-1.5-Flash, validates them, and pushes them here.
+
+- **Source Code:** [Link to your GitHub Profile]
+- **Visualization App:** [Link to your Space]