File size: 11,040 Bytes
ce86b63 c6bae3c ce86b63 c6bae3c ce86b63 c6bae3c ce86b63 c6bae3c ce86b63 c6bae3c ce86b63 c6bae3c ce86b63 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 |
---
library_name: transformers
license: cc-by-nc-4.0
pipeline_tag: text-generation
tags:
- text-to-sql
- reinforcement-learning
---
# SLM-SQL: An Exploration of Small Language Models for Text-to-SQL
### Important Links
📖[Arxiv Paper](https://arxiv.org/abs/2507.22478) |
\ud83d\udcbb[GitHub](https://github.com/CycloneBoy/slm_sql) |
\ud83e\udd17[HuggingFace](https://huggingface.co/collections/cycloneboy/slm-sql-688b02f99f958d7a417658dc) |
\ud83e\udd16[ModelScope](https://modelscope.cn/collections/SLM-SQL-624bb6a60e9643) |
## News
+ `July 31, 2025`: Upload model to modelscope and huggingface.
+ `July 30, 2025`: Publish the paper to arxiv
## Introduction
> Large language models (LLMs) have demonstrated strong performance in translating natural language questions into SQL
> queries (Text-to-SQL). In contrast, small language models (SLMs) ranging from 0.5B to 1.5B parameters currently
> underperform on Text-to-SQL tasks due to their limited logical reasoning capabilities. However, SLMs offer inherent
> advantages in inference speed and suitability for edge deployment. To explore their potential in Text-to-SQL
> applications, we leverage recent advancements in post-training techniques. Specifically, we used the open-source
> SynSQL-2.5M dataset to construct two derived datasets: SynSQL-Think-916K for SQL generation and
> SynSQL-Merge-Think-310K
> for SQL merge revision. We then applied supervised fine-tuning and reinforcement learning-based post-training to the
> SLM, followed by inference using a corrective self-consistency approach. Experimental results validate the
> effectiveness
> and generalizability of our method, SLM-SQL. On the BIRD development set, the five evaluated models achieved an
> average
> improvement of 31.4 points. Notably, the 0.5B model reached 56.87% execution accuracy (EX), while the 1.5B model
> achieved 67.08% EX. We will release our dataset, model, and code to github: https://github.com/CycloneBoy/slm_sql.
### Framework
<img src="https://raw.githubusercontent.com/CycloneBoy/slm_sql/main/data/image/slmsql_framework.png" height="500" alt="slmsql_framework">
### Main Results
<img src="https://raw.githubusercontent.com/CycloneBoy/slm_sql/main/data/image/slmsql_bird_result.png" height="500" alt="slm_sql_result">
<img src="https://raw.githubusercontent.com/CycloneBoy/slm_sql/main/data/image/slmsql_bird_main.png" height="500" alt="slmsql_bird_main">
<img src="https://raw.githubusercontent.com/CycloneBoy/slm_sql/main/data/image/slmsql_spider_main.png" height="500" alt="slmsql_spider_main">
Performance Comparison of different Text-to-SQL methods on BIRD dev and test dataset.
<img src="https://raw.githubusercontent.com/CycloneBoy/slm_sql/main/data/image/slmsql_ablation_study.png" height="300" alt="slmsql_ablation_study">
## Usage
Here's how to use the model for Text-to-SQL generation.
```python
import torch
from transformers import AutoTokenizer, AutoModelForCausalLM
model_id = "cycloneboy/SLM-SQL-0.5B" # Or choose another model from the table above
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(
model_id,
torch_dtype=torch.bfloat16,
device_map="auto",
)
# Example query
question = "How many members are there in the department of 'Sales'?"
# The chat template is critical for proper inference as the model is instruction-tuned.
messages = [
{"role": "system", "content": "You are an AI programming assistant, utilizing the Deepseek Coder model, developed by Deepseek Company, and you only answer questions related to computer science. For politically sensitive questions, security and privacy issues, and other non-computer science questions, you will refuse to answer
"},
{"role": "user", "content": f"### Instruction:
Generate a SQL query for the following question:
{question}
"},
]
# Apply the chat template to get the formatted prompt string
prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
# Tokenize and generate
input_ids = tokenizer(prompt, return_tensors="pt").input_ids.to(model.device)
outputs = model.generate(input_ids, max_new_tokens=256, do_sample=True, temperature=0.01, top_p=0.95)
# Decode the generated text, skipping special tokens
generated_text = tokenizer.decode(outputs[0], skip_special_tokens=True)
# Extract only the model's response based on the chat template's structure
# The response should start after "### Response:
" and end before "<|EOT|>
"
response_start = generated_text.find("### Response:
")
if response_start != -1:
response_content = generated_text[response_start + len("### Response:
"):]
response_end = response_content.find("<|EOT|>")
if response_end != -1:
sql_query = response_content[:response_end].strip()
print(f"Generated SQL: {sql_query}")
else:
print(f"Generated text (full): {response_content.strip()}")
else:
print(f"Generated text (full): {generated_text.strip()}")
```
## Model
| **Model** | Base Model | Train Method | Modelscope | HuggingFace |
|------------------------------------------|------------------------------|--------------|---------------------------------------------------------------------------------------------------|----------------------------------------------------------------------------------------------|
| SLM-SQL-Base-0.5B | Qwen2.5-Coder-0.5B-Instruct | SFT | [\ud83e\udd16 Modelscope](https://modelscope.cn/models/cycloneboy/SLM-SQL-Base-0.5B) | [\ud83e\udd17 HuggingFace](https://huggingface.co/cycloneboy/SLM-SQL-Base-0.5B) |
| SLM-SQL-0.5B | Qwen2.5-Coder-0.5B-Instruct | SFT + GRPO | [\ud83e\udd16 Modelscope](https://modelscope.cn/models/cycloneboy/SLM-SQL-0.5B) | [\ud83e\udd17 HuggingFace](https://huggingface.co/cycloneboy/SLM-SQL-0.5B) |
| CscSQL-Merge-Qwen2.5-Coder-0.5B-Instruct | Qwen2.5-Coder-0.5B-Instruct | SFT + GRPO | [\ud83e\udd16 Modelscope](https://modelscope.cn/models/cycloneboy/CscSQL-Merge-Qwen2.5-Coder-0.5B-Instruct) | [\ud83e\udd17 HuggingFace](https://huggingface.co/cycloneboy/CscSQL-Merge-Qwen2.5-Coder-0.5B-Instruct) |
| SLM-SQL-Base-1.5B | Qwen2.5-Coder-1.5B-Instruct | SFT | [\ud83e\udd16 Modelscope](https://modelscope.cn/models/cycloneboy/SLM-SQL-Base-1.5B) | [\ud83e\udd17 HuggingFace](https://huggingface.co/cycloneboy/SLM-SQL-Base-1.5B) |
| SLM-SQL-1.5B | Qwen2.5-Coder-1.5B-Instruct | SFT + GRPO | [\ud83e\udd16 Modelscope](https://modelscope.cn/models/cycloneboy/SLM-SQL-1.5B) | [\ud83e\udd17 HuggingFace](https://huggingface.co/cycloneboy/SLM-SQL-1.5B) |
| CscSQL-Merge-Qwen2.5-Coder-1.5B-Instruct | Qwen2.5-Coder-1.5B-Instruct | SFT + GRPO | [\ud83e\udd16 Modelscope](https://modelscope.cn/models/cycloneboy/CscSQL-Merge-Qwen2.5-Coder-1.5B-Instruct) | [\ud83e\udd17 HuggingFace](https://huggingface.co/cycloneboy/CscSQL-Merge-Qwen2.5-Coder-1.5B-Instruct) |
| SLM-SQL-Base-0.6B | Qwen3-0.6B | SFT | [\ud83e\udd16 Modelscope](https://modelscope.cn/models/cycloneboy/SLM-SQL-Base-0.6B) | [\ud83e\udd17 HuggingFace](https://huggingface.co/cycloneboy/SLM-SQL-Base-0.6B) |
| SLM-SQL-0.6B | Qwen3-0.6B | SFT + GRPO | [\ud83e\udd16 Modelscope](https://modelscope.cn/models/cycloneboy/SLM-SQL-0.6B) | [\ud83e\udd17 HuggingFace](https://huggingface.co/cycloneboy/SLM-SQL-0.6B) |
| SLM-SQL-Base-1.3B | deepseek-coder-1.3b-instruct | SFT | [\ud83e\udd16 Modelscope](https://modelscope.cn/models/cycloneboy/SLM-SQL-Base-1.3B ) | [\ud83e\udd17 HuggingFace](https://huggingface.co/cycloneboy/SLM-SQL-Base-1.3B ) |
| SLM-SQL-1.3B | deepseek-coder-1.3b-instruct | SFT + GRPO | [\ud83e\udd16 Modelscope](https://modelscope.cn/models/cycloneboy/SLM-SQL-1.3B ) | [\ud83e\udd17 HuggingFace](https://huggingface.co/cycloneboy/SLM-SQL-1.3B ) |
| SLM-SQL-Base-1B | Llama-3.2-1B-Instruct | SFT | [\ud83e\udd16 Modelscope](https://modelscope.cn/models/cycloneboy/SLM-SQL-Base-1B ) | [\ud83e\udd17 HuggingFace](https://huggingface.co/cycloneboy/SLM-SQL-Base-1B ) |
## Dataset
| **Dataset** | Modelscope | HuggingFace |
|----------------------------|------------------------------------------------------------------------------------|--------------------------------------------------------------------------------------|
| SynsQL-Think-916k | [\ud83e\udd16 Modelscope](https://modelscope.cn/datasets/cycloneboy/SynsQL-Think-916k) | [\ud83e\udd17 HuggingFace](https://huggingface.co/datasets/cycloneboy/SynsQL-Think-916k) |
| SynsQL-Merge-Think-310k | [\ud83e\udd16 Modelscope](https://modelscope.cn/datasets/cycloneboy/SynsQL-Merge-Think-310k) | [\ud83e\udd17 HuggingFace](https://huggingface.co/datasets/cycloneboy/SynsQL-Merge-Think-310k) |
| bird train and dev dataset | [\ud83e\udd16 Modelscope](https://modelscope.cn/datasets/cycloneboy/bird_train) | [\ud83e\udd17 HuggingFace](https://huggingface.co/datasets/cycloneboy/bird_train) |
## TODO
- [ ] Release inference code
- [ ] Upload Model
- [ ] Release training code
- [ ] Fix bug
- [ ] Update doc
## Thanks to the following projects
- [csc_sql](https://github.com/CycloneBoy/csc_sql)
- [open-r1](https://github.com/huggingface/open-r1)
- [OmniSQL](https://github.com/RUCKBReasoning/OmniSQL)
## Citation
```bibtex
@misc{sheng2025slmsqlexplorationsmalllanguage,
title={SLM-SQL: An Exploration of Small Language Models for Text-to-SQL},
author={Lei Sheng and Shuai-Shuai Xu},
year={2025},
eprint={2507.22478},
archivePrefix={arXiv},
primaryClass={cs.CL},
url={https://arxiv.org/abs/2507.22478},
}
@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},
}
``` |