Table Question Answering on WikiTQ (Accuracy, Average Improvement, Cohen's d)
77.2AccuracyDataFactory
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| DataFactoryBackbone LLM=Gemini 2.5 Flash2026.03 | 77.2 | 72.8 | 8.71 | |
| TabSQLifyBackbone LLM=GPT-4o mini2026.03 | 74 | 69.6 | 5.24 | |
| DataFactoryBackbone LLM=Deepseek V32026.03 | 73.5 | 72.8 | 8.71 | |
| StructGPTBackbone LLM=Gemini 2.5 Flash2026.03 | 70.9 | 62.1 | 8.03 | |
| ReAcTableBackbone LLM=Gemini 2.5 Flash2026.03 | 70.4 | 61.7 | 4.61 | |
| MACTBackbone LLM=Deepseek V32026.03 | 68.6 | 66.1 | 4.8 | |
| AutoPrepBackbone LLM=Gemini 2.5 Flash2026.03 | 68.2 | 67.2 | 2.57 | |
| TabSQLifyBackbone LLM=Deepseek V32026.03 | 68 | 69.6 | 5.24 | |
| DataFactoryBackbone LLM=GPT-4o mini2026.03 | 67.7 | 72.8 | 8.71 | |
| AutoPrepBackbone LLM=Deepseek V32026.03 | 67.5 | 67.2 | 2.57 | |
| MACTBackbone LLM=Gemini 2.5 Flash2026.03 | 67.3 | 66.1 | 4.8 | |
| TabSQLifyBackbone LLM=Gemini 2.5 Flash2026.03 | 66.7 | 69.6 | 5.24 | |
| AutoPrepBackbone LLM=GPT-4o mini2026.03 | 65.9 | 67.2 | 2.57 | |
| StructGPTBackbone LLM=Deepseek V32026.03 | 65.5 | 62.1 | 8.03 | |
| MACTBackbone LLM=GPT-4o mini2026.03 | 62.3 | 66.1 | 4.8 | |
| TableCoTBackbone LLM=Deepseek V32026.03 | 60.6 | 45 | 13.65 | |
| ReAcTableBackbone LLM=GPT-4o mini2026.03 | 57.8 | 61.7 | 4.61 | |
| End-to-EndBackbone LLM=Gemini 2.5 Flash2026.03 | 57.1 | 54.6 | 16.59 | |
| ReAcTableBackbone LLM=Deepseek V32026.03 | 56.8 | 61.7 | 4.61 | |
| TAPEX-large2026.03 | 56.2 | 56.2 | 9.76 | |
| End-to-EndBackbone LLM=Deepseek V32026.03 | 55.9 | 54.6 | 16.59 | |
| Few-ShotBackbone LLM=Deepseek V32026.03 | 55.6 | 49.6 | 13.94 | |
| TAPAS-large2026.03 | 50.7 | 50.7 | 7.89 | |
| End-to-EndBackbone LLM=GPT-4o mini2026.03 | 50.7 | 54.6 | 16.59 | |
| Few-ShotBackbone LLM=Gemini 2.5 Flash2026.03 | 50.6 | 49.6 | 13.94 | |
| StructGPTBackbone LLM=GPT-4o mini2026.03 | 49.8 | 62.1 | 8.03 | |
| Few-ShotBackbone LLM=GPT-4o mini2026.03 | 42.7 | 49.6 | 13.94 | |
| TableCoTBackbone LLM=Gemini 2.5 Flash2026.03 | 42.2 | 45 | 13.65 | |
| TableCoTBackbone LLM=GPT-4o mini2026.03 | 32.3 | 45 | 13.65 |