Table-based Question Answering on WikiTableQuestion (official)
65.9Test AccuracyCodex w/ DATER
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Codex w/ DATERLearning Paradigm=LLM based2023.01 | 65.9 | 64.8 | |
| OmniTab w/ DATERLearning Paradigm=Fine-tuning based2023.01 | 62.5 | 62.5 | |
| BinderLearning Paradigm=LLM based2023.01 | 61.9 | 62.6 | |
| OmniTabLearning Paradigm=Fine-tuning based2023.01 | 61.2 | 61.3 | |
| TaCubeLearning Paradigm=Fine-tuning based2023.01 | 60.8 | 60.9 | |
| ReasTAPLearning Paradigm=Fine-tuning based2023.01 | 58.6 | 58.3 | |
| TAPEXLearning Paradigm=Fine-tuning based2023.01 | 57.2 | 58 | |
| TableFormerLearning Paradigm=Fine-tuning based2023.01 | 52.6 | 51.3 | |
| TaPasLearning Paradigm=Fine-tuning based2023.01 | 50.4 | 49.9 | |
| T5-3BLearning Paradigm=Fine-tuning based2023.01 | 49.3 | — | |
| CodexLearning Paradigm=LLM based2023.01 | 47.6 | 49.3 | |
| IterativeSearchLearning Paradigm=Fine-tuning based2023.01 | 44.7 | 43.1 | |
| LatentAlignmentLearning Paradigm=Fine-tuning based2023.01 | 44.5 | 43.7 | |
| MeRLLearning Paradigm=Fine-tuning based2023.01 | 44.1 | 43.2 | |
| MAPOLearning Paradigm=Fine-tuning based2023.01 | 43.8 | 42.7 |