Table Question Answering on ROBUT-WTQ Synonym Replacement (test)
83.6Accuracy (Pre-perturbation)TAPEX
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| TAPEXEvaluation Protocol=Fine-tuned2023.06 | 83.6 | 68.8 | 79.5 | |
| OmniTabEvaluation Protocol=Fine-tuned2023.06 | 82.3 | 70.7 | 82 | |
| TAPASEvaluation Protocol=Fine-tuned2023.06 | 81.2 | 62.2 | 73.1 | |
| TableFormerEvaluation Protocol=Fine-tuned2023.06 | 80.7 | 64 | 75.8 | |
| GPT-3Evaluation Protocol=Few-shot, Sample Size=2002023.06 | 78.1 | 74.5 | 91.8 | |
| OmniTabEvaluation Protocol=Fine-tuned2023.06 | 58.5 | 54 | 88 | |
| TAPEXEvaluation Protocol=Fine-tuned2023.06 | 54.3 | 48.4 | 84.6 | |
| TableFormerEvaluation Protocol=Fine-tuned2023.06 | 47 | 41.1 | 83.2 | |
| TAPASEvaluation Protocol=Fine-tuned2023.06 | 44.7 | 38.5 | 81.1 | |
| GPT-3Evaluation Protocol=Few-shot2023.06 | 41.7 | 39.9 | 90.7 |