Update README.md
Browse files
README.md
CHANGED
@@ -2,4 +2,39 @@
|
|
2 |
license: apache-2.0
|
3 |
base_model:
|
4 |
- Qwen/Qwen2.5-Coder-7B-Instruct
|
5 |
-
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2 |
license: apache-2.0
|
3 |
base_model:
|
4 |
- Qwen/Qwen2.5-Coder-7B-Instruct
|
5 |
+
---
|
6 |
+
|
7 |
+
## Overview
|
8 |
+
Arctic-Text2SQL-R1-7B is a 7-billion-parameter Text-to-SQL model fine-tuned using Group Relative Policy Optimization (GRPO) with a simple execution-based reward signal. It converts natural language questions into executable SQL queries.
|
9 |
+
|
10 |
+
## Key Features
|
11 |
+
|
12 |
+
- **Lightweight RL formulation**: Uses only execution correctness and syntax validity as rewards.
|
13 |
+
- **State-of-the-art performance**: Achieves 68.9% execution accuracy on BIRD-dev and 68.5% on BIRD-test, with an average of 57.2% across six benchmarks (BIRD, Spider, Spider2.0, Spider-DK, EHRSQL, ScienceBenchmark)
|
14 |
+
- **Efficiency**: Outperforms many 70B+ models with only 7B parameters.
|
15 |
+
|
16 |
+
## Intended Use
|
17 |
+
|
18 |
+
This model is designed for:
|
19 |
+
|
20 |
+
- Interactive natural language interfaces to relational databases.
|
21 |
+
- Data analytics tools enabling non-technical users to query databases.
|
22 |
+
|
23 |
+
### Not intended for:
|
24 |
+
- Generation of non-SQL text or free-form natural language tasks.
|
25 |
+
- Production systems without validation, especially in safety-critical domains.
|
26 |
+
|
27 |
+
| Benchmark | Dev/Test Accuracy |
|
28 |
+
| ---------------- | ----------------- |
|
29 |
+
| BIRD-dev | 68.9% |
|
30 |
+
| BIRD-test | 68.5% |
|
31 |
+
| Spider-test | 88.8% |
|
32 |
+
| Spider2.0-DK | 15.6% |
|
33 |
+
| EHRSQL | 36.7% |
|
34 |
+
| ScienceBenchmark | 51.8% |
|
35 |
+
| **Average** | **57.2%** |
|
36 |
+
|
37 |
+
|
38 |
+
## Ethical Considerations
|
39 |
+
- Avoid using for private or sensitive data without proper oversight.
|
40 |
+
- Validate generated SQL to prevent data leakage or unauthorized access.
|