Text-to-SQL on EHRSQL (Gre/Maj Metrics)
56.1Gre ScoreDataFlow-Text2SQL-90K
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| DataFlow-Text2SQL-90KBackbone=Qwen2.5-Coder-7B-Instruct2025.12 | 56.1 | 58.7 | |
| DataFlow-Text2SQL-90KBackbone=Meta-Llama-3.1-8B-Instruct2025.12 | 55.5 | 56.3 | |
| GPT-4oAccess=Closed-source2025.12 | 44.9 | 45.5 | |
| GPT-4-TurboAccess=Closed-source2025.12 | 43.1 | 44.8 | |
| GPT-4o-miniAccess=Closed-source2025.12 | 37.9 | 43.1 | |
| SynSQL(2.5M)Backbone=Meta-Llama-3.1-8B-Instruct2025.12 | 35.1 | 37 | |
| SynSQL(2.5M)Backbone=Qwen2.5-Coder-7B-Instruct2025.12 | 34.9 | 40 | |
| SynSQL(50K)Backbone=Meta-Llama-3.1-8B-Instruct2025.12 | 33.6 | 40.8 | |
| SynSQL(50K)Backbone=Qwen2.5-Coder-7B-Instruct2025.12 | 33.1 | 44.1 | |
| SynSQL(90K)Backbone=Meta-Llama-3.1-8B-Instruct2025.12 | 31.8 | 38 | |
| SynSQL(90K)Backbone=Qwen2.5-Coder-7B-Instruct2025.12 | 31.4 | 41.4 | |
| DeepSeek-Coder-7B-InstructAccess=Open-source2025.12 | 28.6 | 33.9 | |
| Spider+BIRD+DataFlow-Text2SQL-90KBackbone=Meta-Llama-3.1-8B-Instruct2025.12 | 28.4 | 36.5 | |
| DataFlow-Text2SQL-50KBackbone=Meta-Llama-3.1-8B-Instruct2025.12 | 28 | 36.4 | |
| Spider+BIRD+DataFlow-Text2SQL-90KBackbone=Qwen2.5-Coder-7B-Instruct2025.12 | 27.9 | 39.8 | |
| DataFlow-Text2SQL-50KBackbone=Qwen2.5-Coder-7B-Instruct2025.12 | 27.8 | 39.4 | |
| Meta-Llama-3.1-8B-InstructAccess=Open-source2025.12 | 24.6 | 33.7 | |
| Qwen2.5-Coder-7B-InstructAccess=Open-source2025.12 | 24.3 | 36.9 | |
| OpenCoder-8B-InstructAccess=Open-source2025.12 | 21.9 | 29.9 | |
| Qwen2.5-7B-InstructAccess=Open-source2025.12 | 20.9 | 32.1 | |
| Granite-3.1-8B-InstructAccess=Open-source2025.12 | 19.6 | 32.3 | |
| Granite-8B-Code-InstructAccess=Open-source2025.12 | 16 | 22.6 |