Table Question Answering on ROBUT-SQA Natural Language Query Perturbations
68.5Pre-perturbation AccuracyTAPEX
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| TAPEXPerturbation Type=Word-Level Paraphrase2023.06 | 68.5 | 63.1 | 5.4 | 86.9 | |
| TAPEXPerturbation Type=Sentence-Level Paraphrase2023.06 | 68 | 61.9 | 6.1 | 86 | |
| TAPASPerturbation Type=Word-Level Paraphrase2023.06 | 63.6 | 57.7 | 5.9 | 86.1 | |
| TAPASPerturbation Type=Sentence-Level Paraphrase2023.06 | 62.8 | 57.5 | 5.3 | 85.7 | |
| GPT-3few-shot=true, Perturbation Type=Word-Level Paraphrase2023.06 | 50.4 | 49.8 | 0.6 | 95.2 | |
| GPT-3few-shot=true, Perturbation Type=Sentence-Level Paraphrase2023.06 | 49.8 | 49.5 | 0.3 | 96.3 |