Snowflake/Arctic-Text2SQL-R1-7B

1---2license: apache-2.03base_model:4- Qwen/Qwen2.5-Coder-7B-Instruct5---6 7## Overview8Arctic-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 10Read more in our paper: [Arctic-Text2SQL-R1: Simple Rewards, Strong Reasoning in Text-to-SQL](https://arxiv.org/abs/2505.20315).11 12## Key Features13 14- **Lightweight RL formulation**: Uses only execution correctness and syntax validity as rewards.15- **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)16- **Efficiency**: Outperforms many 70B+ models with only 7B parameters.17 18## Intended Use19 20This model is designed for:21 22- Interactive natural language interfaces to relational databases.23- Data analytics tools enabling non-technical users to query databases.24 25### Not intended for:26- Generation of non-SQL text or free-form natural language tasks.27- Production systems without validation, especially in safety-critical domains.28 29| Benchmark        | Dev/Test Accuracy |30| ---------------- | ----------------- |31| BIRD-dev         | 68.9%             |32| BIRD-test        | 68.5%             |33| Spider-test      | 88.8%             |34| Spider2.0-DK     | 15.6%             |35| EHRSQL           | 36.7%             |36| ScienceBenchmark | 51.8%             |37| **Average**      | **57.2%**         |38 39 40## Ethical Considerations41- Avoid using for private or sensitive data without proper oversight.42- Validate generated SQL to prevent data leakage or unauthorized access.