Update README.md

README.md

@@ -8,3 +8,32 @@ This is a preview of our up and comming synthetic dataset: Rephrased2JSON.
 
 ## Dataset Summary
 
+Rephrased2JSON-Preview is a synthetically generated, cross-domain dataset designed to train large language models on structured information. The dataset transforms unstructured natural language into semantic JSON representations capturing key facts, named entities, relationships, abstract concepts, and quantitative metrics.
+
+This preview release contains ~20,000 high-density extraction records generated locally using Liquid AI's LFM2.5-1.2B-Instruct.
+
+## Schema structure
+
+```
+{
+  "doc_id": "dclm_14028",
+  "text": "... [Source passage truncated up to 2,048 tokens] ...",
+  "text_len": 1102,
+  "text_from": "HuggingFaceTB/dclm-edu",
+  "results": [
+    {
+      "generated_output": "{\n  \"subject\": \"...\",\n  \"key_facts\": [...],\n  \"entities\": [...],\n  \"concepts\": [...],\n  \"quantitative_metrics\": [\n    {\"metric\": \"context_window\", \"value\": 4096}\n  ]\n}",
+      "len": 384,
+      "valid": true,
+      "grammar_constrained": true,
+      "type_profile": {
+        "numbers": 4,
+        "strings": 28,
+        "arrays": 4,
+        "objects": 6,
+        "booleans": 0
+      }
+    }
+  ]
+}
+```