Text-to-SQL on Spider-Syn
83.1Execution Accuracy (EX)FineStep-4B
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| FineStep-4BModel Size=4B2026.05 | 83.1 | — | — | — | — | 75.2 | |
| FineStepModel Size=4B2026.05 | 83.1 | — | — | — | — | 75.2 | |
| EXPO-SQLDetail=RL, Base Model=Qwen2.5-Coder-7B2026.04 | 83.1 | — | — | — | — | — | |
| FineStep-8BModel Size=8B2026.05 | 81.4 | — | — | — | — | 71.6 | |
| FineStepModel Size=8B2026.05 | 81.4 | — | — | — | — | 71.6 | |
| MTIR-SQL + Qwen3-14BBackbone=Qwen3-14B2026.05 | 81 | — | — | — | — | — | |
| FineStepModel Size=30B-A3B2026.05 | 80.8 | — | — | — | — | 72.6 | |
| GIGPOModel Size=8B2026.05 | 80.4 | — | — | — | — | 71.7 | |
| GIGPOModel Size=4B2026.05 | 80.2 | — | — | — | — | 71.7 | |
| GSPOModel Size=4B2026.05 | 80.1 | — | — | — | — | 72.1 | |
| DAPOModel Size=4B2026.05 | 80 | — | — | — | — | 72.5 | |
| GSPOModel Size=8B2026.05 | 79.6 | — | — | — | — | 70.3 | |
| GRPOModel Size=30B-A3B2026.05 | 79.6 | — | — | — | — | 70.4 | |
| DAPOModel Size=8B2026.05 | 79.5 | — | — | — | — | 71.1 | |
| GRPOModel Size=8B2026.05 | 79.2 | — | — | — | — | 70 | |
| Reward-SQL + Qwen3-8BDecoding=Maj@82025.05 | 79 | — | — | — | — | — | |
| GRPOModel Size=4B2026.05 | 78.9 | — | — | — | — | 71.2 | |
| DCG-SQLLLM Backbone=GPT-42025.05 | 78.7 | 68.6 | — | — | — | — | |
| ReEx-SQL-7BModel Size=7B2026.05 | 78.7 | — | — | — | — | 72 | |
| Think2SQL-14BModel Scale Category=10-50B2025.04 | 78.6 | — | — | — | — | — | |
| SENSE-13BModel Size=13B2026.05 | 77.6 | — | — | — | — | 70.2 | |
| SQL-o1 + Llama3-8BBackbone=Llama3-8B2026.05 | 77.6 | — | — | — | — | 69.2 | |
| ROUTE + Llama3-8BBackbone=Llama3-8B2026.05 | 77.4 | — | — | — | — | 70.2 | |
| MTIR-SQLModel Scale=8B, Decoding Strategy=greedy2026.06 | 77.2 | — | — | — | — | — | |
| Progress-SQL-14BReward=Progressive2026.06 | 77.1 | — | — | — | — | 69.1 | |
| SFT CodeS-15BModel Size=15B2026.05 | 77 | — | — | — | — | 69.4 | |
| OmniSQL-7BReward=–2026.06 | 77 | — | — | — | — | 70.2 | |
| SFT CodeS-7BReward=–2026.06 | 76.9 | — | — | — | — | 70 | |
| SQL-R1-7B*Model Size=7B2026.05 | 76.7 | — | — | — | — | — | |
| SQL-R1Detail=RL, Base Model=Qwen2.5-Coder-7B2026.04 | 76.7 | — | — | — | — | — | |
| Progress-SQL-7BReward=Progressive2026.06 | 76.6 | — | — | — | — | 68.8 | |
| No-TrainModel Size=4B2026.05 | 76.1 | — | — | — | — | 65 | |
| CYCLESQL + GPT-4Backbone=GPT-42026.05 | 76 | — | — | — | — | 66.3 | |
| Arctic-SQL + Qwen2.5-7BDecoding=Maj@82025.05 | 76 | — | — | — | — | — | |
| Reward-SQL + Qwen3-8BDecoding=Greedy2025.05 | 75.7 | — | — | — | — | — | |
| No-TrainModel Size=8B2026.05 | 75.4 | — | — | — | — | 62 | |
| No-TrainModel Size=30B-A3B2026.05 | 75.4 | — | — | — | — | 61.2 | |
| SQL-R1 + Qwen2.5-7BDecoding=Maj@82025.05 | 75.4 | — | — | — | — | — | |
| Arctic-SQL + Qwen2.5-7BDecoding=Greedy2025.05 | 75.2 | — | — | — | — | — | |
| Think2SQL-4BModel Scale Category=< 5B2025.04 | 75 | — | — | — | — | — | |
| Think2SQL-8BModel Scale Category=5-10B2025.04 | 74.9 | — | — | — | — | — | |
| KE-RLLLM=Qwen2.5-Coder-7B2026.05 | 74.83 | — | — | — | — | — | |
| SQL-PaLM + PaLM2Backbone=PaLM22026.05 | 74.6 | — | — | — | — | — | |
| ASTRESLLM Backbone=GPT-42025.05 | 74.4 | 61.3 | — | — | — | — | |
| FastRAText+GPT-4Backbone=GPT-42026.05 | 74.4 | — | — | — | — | — | |
| DCG-SQLLLM Backbone=GPT-3.5-Turbo2025.05 | 74.2 | 61.6 | — | — | — | — | |
| OmniSQL + Qwen2.5-7BDecoding=Maj@82025.05 | 73.9 | — | — | — | — | — | |
| ACE-SQLDecoding Strategy=greedy2026.06 | 73.7 | — | — | — | — | — | |
| SQL-R1-7BReward=One-shot2026.06 | 73.7 | — | — | — | — | 58 | |
| SkyRL-SQL-7BReward=One-shot2026.06 | 73.7 | — | — | — | — | — | |
| SQL-Trail-7BReward=One-shot2026.06 | 73.7 | — | — | — | — | 65.1 | |
| DCG-SQLLLM Backbone=Deepseek-Coder-33B-Instruct2025.05 | 73.6 | 62.6 | — | — | — | — | |
| ACT-SQLLLM Backbone=GPT-42025.05 | 73.1 | 48.5 | — | — | — | — | |
| SQL-R1 + Qwen2.5-7BDecoding=Greedy2025.05 | 72.6 | — | — | — | — | — | |
| SENSE-7BDetail=SFT, Base Model=Qwen2.5-Coder-7B2026.04 | 72.6 | — | — | — | — | — | |
| MAC-SQLDetail=Prompt, Base Model=GPT-42026.04 | 72.5 | — | — | — | — | — | |
| Qwen2.5-Coder-14B-InstructReward=–2026.06 | 72.3 | — | — | — | — | 65.6 | |
| OmniSQL-32BModel Size=32B2026.05 | 72.1 | — | — | — | — | — | |
| KE-ICLLLM=Gemini-Pro-1.52026.05 | 71.76 | — | — | — | — | — | |
| CodeSLLM=Gemini-Pro-1.5, Setting=ICL2026.05 | 71.17 | — | — | — | — | — | |
| Qwen2.5-Coder-7B-InstructReward=–2026.06 | 70.9 | — | — | — | — | 62.3 | |
| OmniSQL + Qwen2.5-7BDecoding=Greedy2025.05 | 70.7 | — | — | — | — | — | |
| DCG-SQLLLM Backbone=Deepseek-Coder-6.7B-Instruct2025.05 | 70.5 | 59.9 | — | — | — | — | |
| ACT-SQLLLM Backbone=GPT-3.5-Turbo2025.05 | 70.4 | 44.4 | — | — | — | — | |
| Qwen2.5-Coder-7BDetail=-, Base Model=Qwen2.5-Coder-7B2026.04 | 70.2 | — | — | — | — | — | |
| DCG-SQLLLM Backbone=Llama-3.1-8B-Instruct2025.05 | 70.1 | 59.7 | — | — | — | — | |
| OmniSQL-7BParameters=7B2026.01 | 69.7 | — | — | — | 61.4 | — | |
| OmniSQLModel Scale=7B, Decoding Strategy=greedy2026.06 | 69.7 | — | — | — | — | — | |
| OmniSQL-7BDetail=SFT, Base Model=Qwen2.5-Coder-7B2026.04 | 69.6 | — | — | — | — | — | |
| DAIL-SQLLLM=Gemini-Pro-1.5, Setting=ICL2026.05 | 69.5 | — | — | — | — | — | |
| Reasoning-SQL*2026.05 | 69.3 | — | — | — | — | — | |
| Reasoning-SQL-7BReward=One-shot2026.06 | 69.3 | — | — | — | — | — | |
| Reasoning-SQLDetail=RL, Base Model=Qwen2.5-Coder-7B2026.04 | 69.3 | — | — | — | — | — | |
| EVOLSQL-7BParameters=7B2026.01 | 69.2 | — | — | — | 61.9 | — | |
| OmniLLM=Qwen2.5-Coder-7B, Setting=RL2026.05 | 69.17 | — | — | — | — | — | |
| ACT-SQLLLM Backbone=Deepseek-Coder-33B-Instruct2025.05 | 69.1 | 49.3 | — | — | — | — | |
| ASTRESLLM Backbone=Deepseek-Coder-33B-Instruct2025.05 | 68.7 | 49.5 | — | — | — | — | |
| ACT-SQLDetail=Prompt, Base Model=GPT-42026.04 | 67.9 | — | — | — | — | — | |
| SQL-GENLLM=Qwen2.5-Coder-7B, Setting=RL2026.05 | 67.8 | — | — | — | — | — | |
| KE-ICLLLM=Qwen2.5-Coder-7B2026.05 | 66.98 | — | — | — | — | — | |
| ASTRESLLM Backbone=GPT-3.5-Turbo2025.05 | 66.9 | 51.2 | — | — | — | — | |
| Qwen2.5-Coder-7B-InstructDecoding=Maj@82025.05 | 66.9 | — | — | — | — | — | |
| ACT-SQLLLM Backbone=Llama-3.1-8B-Instruct2025.05 | 66.6 | 53.7 | — | — | — | — | |
| KE-ICLLLM=GPT-4o2026.05 | 66.05 | — | — | — | — | — | |
| KE-RLLLM=Deepseek-Coder-7B2026.05 | 66.04 | — | — | — | — | — | |
| SFT (BIRD+Spider)Training Data=BIRD+Spider2026.01 | 65.9 | — | — | — | 55.8 | — | |
| CodeSLLM=GPT-4o, Setting=ICL2026.05 | 65.86 | — | — | — | — | — | |
| DAIL-SQLLLM=Qwen2.5-Coder-7B, Setting=ICL2026.05 | 65.43 | — | — | — | — | — | |
| DIN-SQLLLM Backbone=GPT-3.5-Turbo2025.05 | 65.3 | 34.7 | — | — | — | — | |
| DeepSeek-V3 (671B MoE)Decoding=Maj@82025.05 | 65.1 | — | — | — | — | — | |
| CodeSLLM=Qwen2.5-Coder-7B, Setting=ICL2026.05 | 64.77 | — | — | — | — | — | |
| DeepSeek-V3 (671B MoE)Decoding=Greedy2025.05 | 64.4 | — | — | — | — | — | |
| Qwen2.5-72B-InstructDecoding=Maj@82025.05 | 64.3 | — | — | — | — | — | |
| Qwen2.5-72B-InstructDecoding=Greedy2025.05 | 64.1 | — | — | — | — | — | |
| KE-ICLLLM=Deepseek-Coder-7B2026.05 | 64 | — | — | — | — | — | |
| GPT-4-TurboDecoding=Maj@82025.05 | 63.5 | — | — | — | — | — | |
| DAIL-SQLLLM=GPT-4o, Setting=ICL2026.05 | 63.44 | — | — | — | — | — | |
| ACT-SQLLLM Backbone=Deepseek-Coder-6.7B-Instruct2025.05 | 63.2 | 44.6 | — | — | — | — | |
| Base ModelBackbone=Qwen2.5-Coder-7B-Instruct2026.01 | 63.1 | — | — | — | 53.4 | — | |
| Qwen2.5-Coder-7B-InstructDecoding=Greedy2025.05 | 63.1 | — | — | — | — | — |