# Table Question Answering on ROBUT-SQA Table Content Perturbations

Source: [Table Question Answering on ROBUT-SQA Table Content Perturbations](<https://www.sota2.com/research/sota/table-question-answering-on-robut-sqa-table-content-perturbations>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-02-26T05:58:11.142Z

Evaluation of model performance on the ROBUT-SQA Table Content Perturbations dataset variant for Table Question Answering, focusing on pre/post-perturbation accuracy and robustness metrics.

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

| Scope | Recorded value |
| --- | --- |
| Task | Table Question Answering |
| Dataset (including any stated split) | ROBUT-SQA Table Content Perturbations |
| Metric | Pre-perturbation Accuracy |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 15 of 15 |
| Coverage | All results currently returned for this benchmark |

Evaluation settings may differ between results. The original papers and recorded configurations are needed to establish comparability. A shared metric name alone does not establish an identical protocol.

## Metrics

- [Pre-perturbation Accuracy](<https://www.sota2.com/research/sota/table-question-answering-on-robut-sqa-table-content-perturbations.md?metric=Pre-perturbation%20Accuracy>): Higher is better; unit: PERCENT.

- [Post-perturbation Accuracy](<https://www.sota2.com/research/sota/table-question-answering-on-robut-sqa-table-content-perturbations.md?metric=Post-perturbation%20Accuracy>): Higher is better; unit: PERCENT.

- [Accuracy Drop](<https://www.sota2.com/research/sota/table-question-answering-on-robut-sqa-table-content-perturbations.md?metric=Accuracy%20Drop>): Lower is better; unit: PERCENT.

- [Robustness Accuracy](<https://www.sota2.com/research/sota/table-question-answering-on-robut-sqa-table-content-perturbations.md?metric=Robustness%20Accuracy>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Pre-perturbation Accuracy | Result date | Report paper |
| --- | --- | --- | --- |
| TAPEX | 68.3 | Not recorded | 68fa6fdc1e209131cea47b48 |
| TAPEX | 67.3 | Not recorded | 68fa6fdc1e209131cea47b48 |
| TAPEX | 66.8 | Not recorded | 68fa6fdc1e209131cea47b48 |
| TAPEX | 66.7 | Not recorded | 68fa6fdc1e209131cea47b48 |
| TAPAS | 65.2 | Not recorded | 68fa6fdc1e209131cea47b48 |
| TAPEX | 64.1 | Not recorded | 68fa6fdc1e209131cea47b48 |
| TAPAS | 63.3 | Not recorded | 68fa6fdc1e209131cea47b48 |
| TAPAS | 62.7 | Not recorded | 68fa6fdc1e209131cea47b48 |
| TAPAS | 62.4 | Not recorded | 68fa6fdc1e209131cea47b48 |
| TAPAS | 61.3 | Not recorded | 68fa6fdc1e209131cea47b48 |
| GPT-3 | 51.3 | Not recorded | 68fa6fdc1e209131cea47b48 |
| GPT-3 | 50.2 | Not recorded | 68fa6fdc1e209131cea47b48 |
| GPT-3 | 49.5 | Not recorded | 68fa6fdc1e209131cea47b48 |
| GPT-3 | 49.5 | Not recorded | 68fa6fdc1e209131cea47b48 |
| GPT-3 | 49.2 | Not recorded | 68fa6fdc1e209131cea47b48 |

### TAPEX

Pre-perturbation Accuracy: 68.3. Result ID: 6990ef833a6c314849046854.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6fdc1e209131cea47b48.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6bfe1e209131cea31836.md>)

[Implementation](<https://github.com/microsoft/Table-Pretraining>)

Recorded evaluation configuration: {"Perturbation Type":"Column Masking"}. This record does not assert independent reproduction.

### TAPEX

Pre-perturbation Accuracy: 67.3. Result ID: 6990ef833a6c314849046855.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6fdc1e209131cea47b48.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6bfe1e209131cea31836.md>)

[Implementation](<https://github.com/microsoft/Table-Pretraining>)

Recorded evaluation configuration: {"Perturbation Type":"Column Adding"}. This record does not assert independent reproduction.

### TAPEX

Pre-perturbation Accuracy: 66.8. Result ID: 6990ef833a6c314849046853.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6fdc1e209131cea47b48.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6bfe1e209131cea31836.md>)

[Implementation](<https://github.com/microsoft/Table-Pretraining>)

Recorded evaluation configuration: {"Perturbation Type":"Column Extension"}. This record does not assert independent reproduction.

### TAPEX

Pre-perturbation Accuracy: 66.7. Result ID: 6990ef833a6c314849046852.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6fdc1e209131cea47b48.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6bfe1e209131cea31836.md>)

[Implementation](<https://github.com/microsoft/Table-Pretraining>)

Recorded evaluation configuration: {"Perturbation Type":"Column Order Shuffling"}. This record does not assert independent reproduction.

### TAPAS

Pre-perturbation Accuracy: 65.2. Result ID: 6990ef833a6c31484904684f.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6fdc1e209131cea47b48.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6c521e209131cea333e0.md>)

[Implementation](<https://github.com/google-research/tapas>)

Recorded evaluation configuration: {"Perturbation Type":"Column Masking"}. This record does not assert independent reproduction.

### TAPEX

Pre-perturbation Accuracy: 64.1. Result ID: 6990ef833a6c314849046851.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6fdc1e209131cea47b48.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6bfe1e209131cea31836.md>)

[Implementation](<https://github.com/microsoft/Table-Pretraining>)

Recorded evaluation configuration: {"Perturbation Type":"Row Order Shuffling"}. This record does not assert independent reproduction.

### TAPAS

Pre-perturbation Accuracy: 63.3. Result ID: 6990ef833a6c31484904684c.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6fdc1e209131cea47b48.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6c521e209131cea333e0.md>)

[Implementation](<https://github.com/google-research/tapas>)

Recorded evaluation configuration: {"Perturbation Type":"Row Order Shuffling"}. This record does not assert independent reproduction.

### TAPAS

Pre-perturbation Accuracy: 62.7. Result ID: 6990ef833a6c314849046850.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6fdc1e209131cea47b48.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6c521e209131cea333e0.md>)

[Implementation](<https://github.com/google-research/tapas>)

Recorded evaluation configuration: {"Perturbation Type":"Column Adding"}. This record does not assert independent reproduction.

### TAPAS

Pre-perturbation Accuracy: 62.4. Result ID: 6990ef833a6c31484904684e.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6fdc1e209131cea47b48.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6c521e209131cea333e0.md>)

[Implementation](<https://github.com/google-research/tapas>)

Recorded evaluation configuration: {"Perturbation Type":"Column Extension"}. This record does not assert independent reproduction.

### TAPAS

Pre-perturbation Accuracy: 61.3. Result ID: 6990ef833a6c31484904684d.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6fdc1e209131cea47b48.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6c521e209131cea333e0.md>)

[Implementation](<https://github.com/google-research/tapas>)

Recorded evaluation configuration: {"Perturbation Type":"Column Order Shuffling"}. This record does not assert independent reproduction.

### GPT-3

Pre-perturbation Accuracy: 51.3. Result ID: 6990ef833a6c314849046859.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6fdc1e209131cea47b48.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa680c1e209131cea19d53.md>)

[Implementation](<https://github.com/openai/gpt-3/blob/master/overlap_frequency.md>)

Recorded evaluation configuration: {"few-shot":"true","Perturbation Type":"Column Masking"}. This record does not assert independent reproduction.

### GPT-3

Pre-perturbation Accuracy: 50.2. Result ID: 6990ef833a6c31484904685a.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6fdc1e209131cea47b48.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa680c1e209131cea19d53.md>)

[Implementation](<https://github.com/openai/gpt-3/blob/master/overlap_frequency.md>)

Recorded evaluation configuration: {"few-shot":"true","Perturbation Type":"Column Adding"}. This record does not assert independent reproduction.

### GPT-3

Pre-perturbation Accuracy: 49.5. Result ID: 6990ef833a6c314849046857.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6fdc1e209131cea47b48.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa680c1e209131cea19d53.md>)

[Implementation](<https://github.com/openai/gpt-3/blob/master/overlap_frequency.md>)

Recorded evaluation configuration: {"few-shot":"true","Perturbation Type":"Column Order Shuffling"}. This record does not assert independent reproduction.

### GPT-3

Pre-perturbation Accuracy: 49.5. Result ID: 6990ef833a6c314849046858.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6fdc1e209131cea47b48.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa680c1e209131cea19d53.md>)

[Implementation](<https://github.com/openai/gpt-3/blob/master/overlap_frequency.md>)

Recorded evaluation configuration: {"few-shot":"true","Perturbation Type":"Column Extension"}. This record does not assert independent reproduction.

### GPT-3

Pre-perturbation Accuracy: 49.2. Result ID: 6990ef833a6c314849046856.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6fdc1e209131cea47b48.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa680c1e209131cea19d53.md>)

[Implementation](<https://github.com/openai/gpt-3/blob/master/overlap_frequency.md>)

Recorded evaluation configuration: {"few-shot":"true","Perturbation Type":"Row Order Shuffling"}. This record does not assert independent reproduction.

Page 1 of 1. This is an index page, not the full collection.