---
pipeline_tag: text-generation
library_name: transformers
license: cc-by-nc-4.0
tags:
- text-to-sql
- reinforcement-learning
- qwen
---

# CSC-SQL: Corrective Self-Consistency in Text-to-SQL via Reinforcement Learning

This repository contains the `CscSQL-Grpo-Qwen2.5-Coder-7B-Instruct` model, a key component of the CSC-SQL framework, as presented in the paper [CSC-SQL: Corrective Self-Consistency in Text-to-SQL via Reinforcement Learning](https://huggingface.co/papers/2505.13271).

## Abstract
Large language models (LLMs) have demonstrated strong capabilities in translating natural language questions about relational databases into SQL queries. In particular, test-time scaling techniques such as Self-Consistency and Self-Correction can enhance SQL generation accuracy by increasing computational effort during inference. However, these methods have notable limitations: Self-Consistency may select suboptimal outputs despite majority votes, while Self-Correction typically addresses only syntactic errors. To leverage the strengths of both approaches, we propose CSC-SQL, a novel method that integrates Self-Consistency and Self-Correction. CSC-SQL selects the two most frequently occurring outputs from parallel sampling and feeds them into a merge revision model for correction. Additionally, we employ the Group Relative Policy Optimization (GRPO) algorithm to fine-tune both the SQL generation and revision models via reinforcement learning, significantly enhancing output quality. Experimental results confirm the effectiveness and generalizability of CSC-SQL. On the BIRD private test set, our 7B model achieves 71.72% execution accuracy, while the 32B model achieves 73.67%.

For more details, refer to the [paper](https://huggingface.co/papers/2505.13271) and the [official GitHub repository](https://github.com/CycloneBoy/csc_sql).

## Framework Overview
![CSC-SQL Framework](https://raw.githubusercontent.com/CycloneBoy/csc_sql/main/data/image/csc_sql_framework.png)

## Code
The official code repository for CSC-SQL is available on GitHub: [https://github.com/CycloneBoy/csc_sql](https://github.com/CycloneBoy/csc_sql)

## Main Results
Performance comparison of different Text-to-SQL methods on BIRD dev and test dataset:
![CSC-SQL Results](https://raw.githubusercontent.com/CycloneBoy/csc_sql/main/data/image/csc_sql_result_main.png)
<img src="https://raw.githubusercontent.com/CycloneBoy/csc_sql/main/data/image/csc_sql_result_main.png" height="500" alt="csc_sql_result main">

## Model Checkpoints
This model is part of a collection of checkpoints related to CSC-SQL, also available on Hugging Face:

| **Model**                     | HuggingFace                                                                                |
|-------------------------------|--------------------------------------------------------------------------------------------|
| CscSQL-Merge-Qwen2.5-Coder-3B-Instruct | [🤗 HuggingFace](https://huggingface.co/cycloneboy/CscSQL-Merge-Qwen2.5-Coder-3B-Instruct) |
| CscSQL-Merge-Qwen2.5-Coder-7B-Instruct | [🤗 HuggingFace](https://huggingface.co/cycloneboy/CscSQL-Merge-Qwen2.5-Coder-7B-Instruct) |
| CscSQL-Grpo-Qwen2.5-Coder-3B-Instruct | [🤗 HuggingFace](https://huggingface.co/cycloneboy/CscSQL-Grpo-Qwen2.5-Coder-3B-Instruct)  |
| CscSQL-Grpo-XiYanSQL-QwenCoder-3B-2502 | [🤗 HuggingFace](https://huggingface.co/cycloneboy/CscSQL-Grpo-XiYanSQL-QwenCoder-3B-2502) |
| CscSQL-Grpo-Qwen2.5-Coder-7B-Instruct | [🤗 HuggingFace](https://huggingface.co/cycloneboy/CscSQL-Grpo-Qwen2.5-Coder-7B-Instruct)  |
| CscSQL-Grpo-XiYanSQL-QwenCoder-7B-2502 | [🤗 HuggingFace](https://huggingface.co/cycloneboy/CscSQL-Grpo-XiYanSQL-QwenCoder-7B-2502) |

## Usage
You can load this model using the `transformers` library. Here's a basic example for inference:

```python
import torch
from transformers import AutoTokenizer, AutoModelForCausalLM, GenerationConfig

model_name = "cycloneboy/CscSQL-Grpo-Qwen2.5-Coder-7B-Instruct"

# Load model and tokenizer
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(
    model_name,
    torch_dtype=torch.bfloat16, # Or torch.float16 depending on your hardware
    device_map="auto"
)
model.eval()

# Example prompt for text-to-SQL generation
# Note: The prompt format might need to align with the model's specific training
# and database schema format for optimal text-to-SQL performance.
prompt = "Translate the following question to SQL: 'What are the names of all employees?'"

# Encode the prompt
input_ids = tokenizer.encode(prompt, return_tensors="pt").to(model.device)

# Set generation configuration based on the model's generation_config.json
generation_config = GenerationConfig(
    bos_token_id=tokenizer.bos_token_id,
    eos_token_id=[tokenizer.eos_token_id, 151643], # Include <|endoftext|> as eos_token_id
    pad_token_id=tokenizer.bos_token_id, # Or use tokenizer.pad_token_id if different
    temperature=0.7,
    max_new_tokens=512,
    do_sample=True,
    top_p=0.8,
    repetition_penalty=1.1,
    top_k=20,
)

# Generate SQL query
output_ids = model.generate(
    input_ids,
    generation_config=generation_config
)

# Decode the generated SQL
generated_sql = tokenizer.decode(output_ids[0], skip_special_tokens=True)
print(generated_sql)

# For detailed usage, including how to integrate with the full CSC-SQL framework
# for improved accuracy via reinforcement learning, please refer to the
# official GitHub repository: https://github.com/CycloneBoy/csc_sql
```

## Citation
If you find this work helpful or inspiring, please feel free to cite it:

```bibtex
@misc{sheng2025cscsqlcorrectiveselfconsistencytexttosql,
      title={CSC-SQL: Corrective Self-Consistency in Text-to-SQL via Reinforcement Learning}, 
      author={Lei Sheng and Shuai-Shuai Xu},
      year={2025},
      eprint={2505.13271},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2505.13271}, 
}
```