Upload folder using huggingface_hub

README.md

@@ -163,7 +163,24 @@ You are an expert SQL query writer. Follow these rules strictly:
 
 ---
 
-## 📊 Evaluation & Capabilities
+## 🏆 Leaderboard & Benchmark Results
+
+The model was empirically benchmarked on **15-table E-Commerce production schemas** and multi-table natural language query benchmarks running locally via ONNX Runtime GenAI on CPU.
+
+### Performance Summary
+| Model Variant / Pipeline | Execution Accuracy (EX) | Execution Validity Rate | Avg CPU Latency | Model Size |
+|---|:---:|:---:|:---:|:---:|
+| **`spcv` Optimized ONNX Pipeline** | **95.00%** (19/20) | **100.00%** (20/20) | **3,570 ms** | **~980 MB** (INT4) |
+| **`spcv` Original Merged PyTorch Base** | **40.00%** (8/20) | **80.00%** (16/20) | **6,596 ms** | **3.08 GB** (FP16) |
+
+### Benchmark Methodology
+1. **Execution Accuracy (EX)**: Evaluates semantic equivalence by populating SQLite in-memory tables with mock enterprise data and verifying if `PRED_SQL` produces identical output rows/columns to `GOLD_SQL`.
+2. **Schema Complexity**: 15 interconnected production tables (`users`, `orders`, `products`, `shipments`, `payments`, `reviews`, `support_tickets`, etc.).
+3. **Query Diversity**: Evaluates `JOIN` depth (up to 4 tables), nested aggregation (`SUM`, `AVG`, `COUNT`), date arithmetic (`datetime('now', '-30 days')`), subqueries (`NOT IN`), and conditional filtering (`CHECK` constraints).
+
+---
+
+## 📊 Capabilities Breakdown
 
 - **Single & Multi-table JOINs**: Correctly resolves foreign keys and table references.
 - **Aggregations & Filtering**: Accurately computes `SUM`, `COUNT`, `AVG`, `GROUP BY`, and `HAVING` clauses.
@@ -174,4 +191,5 @@ You are an expert SQL query writer. Follow these rules strictly:
 
 ## 📄 License & Attribution
 - Base model licensed under **Apache 2.0** by the Qwen Team / Alibaba Cloud.
-- Distributed as part of the **SLMAgents** multi-agent local intelligence suite.
+- Distributed by **spcv** for high-precision local Text-to-SQL intelligence.
+