# Visual Question Answering on VQA (val) with RAD Metrics

Source: [Visual Question Answering on VQA (val) with RAD Metrics](<https://www.sota2.com/research/sota/visual-question-answering-on-vqa-val-with-rad-metrics>)

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

Data updated: 2026-02-26T05:59:45.020Z

Evaluation of Visual Question Answering performance on the VQA validation set, specifically reporting a set of RAD (Robustness and Adversarial Detection) metrics alongside overall accuracy.

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 | Visual Question Answering |
| Dataset (including any stated split) | VQA (val) |
| Metric | RAD (Y/N ← C) |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 4 of 4 |
| 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

- [RAD (Y/N ← C)](<https://www.sota2.com/research/sota/visual-question-answering-on-vqa-val-with-rad-metrics.md?metric=RAD%20(Y%2FN%20%E2%86%90%20C)>): Higher is better; unit: PERCENT.

- [RAD (Y/N ← HM)](<https://www.sota2.com/research/sota/visual-question-answering-on-vqa-val-with-rad-metrics.md?metric=RAD%20(Y%2FN%20%E2%86%90%20HM)>): Higher is better; unit: PERCENT.

- [RAD (Y/N ← WK)](<https://www.sota2.com/research/sota/visual-question-answering-on-vqa-val-with-rad-metrics.md?metric=RAD%20(Y%2FN%20%E2%86%90%20WK)>): Higher is better; unit: PERCENT.

- [RAD (BT)](<https://www.sota2.com/research/sota/visual-question-answering-on-vqa-val-with-rad-metrics.md?metric=RAD%20(BT)>): Higher is better; unit: PERCENT.

- [RAD (Reph)](<https://www.sota2.com/research/sota/visual-question-answering-on-vqa-val-with-rad-metrics.md?metric=RAD%20(Reph)>): Higher is better; unit: PERCENT.

- [RAD (L-ConVQA)](<https://www.sota2.com/research/sota/visual-question-answering-on-vqa-val-with-rad-metrics.md?metric=RAD%20(L-ConVQA)>): Higher is better; unit: PERCENT.

- [RAD (CS-ConVQA)](<https://www.sota2.com/research/sota/visual-question-answering-on-vqa-val-with-rad-metrics.md?metric=RAD%20(CS-ConVQA)>): Higher is better; unit: PERCENT.

- [Accuracy](<https://www.sota2.com/research/sota/visual-question-answering-on-vqa-val-with-rad-metrics.md?metric=Accuracy>): Higher is better; unit: PERCENT.

## Recorded results

| Method | RAD (Y/N ← C) | Result date | Report paper |
| --- | --- | --- | --- |
| VisualBERT | 79.99 | Not recorded | 68fa6ac91e209131cea2a370 |
| BAN | 74.4 | Not recorded | 68fa6ac91e209131cea2a370 |
| BUTD | 67.15 | Not recorded | 68fa6ac91e209131cea2a370 |
| Pythia | 65 | Not recorded | 68fa6ac91e209131cea2a370 |

### VisualBERT

RAD (Y/N ← C): 79.99. Result ID: 699ad13f5698665ec4603bb3.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6ac91e209131cea2a370.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa66391e209131cea0ef1d.md>)

Recorded evaluation configuration: {"Implementation":"Official"}. This record does not assert independent reproduction.

### BAN

RAD (Y/N ← C): 74.4. Result ID: 699ad13f5698665ec4603bb1.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6ac91e209131cea2a370.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa64fd1e209131cea07a2c.md>)

Recorded evaluation configuration: {"Implementation":"Official"}. This record does not assert independent reproduction.

### BUTD

RAD (Y/N ← C): 67.15. Result ID: 699ad13f5698665ec4603bb0.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6ac91e209131cea2a370.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa642b1e209131cea031b7.md>)

[Implementation](<http://www.panderson.me/up-down-attention>)

Recorded evaluation configuration: {"Implementation":"Official"}. This record does not assert independent reproduction.

### Pythia

RAD (Y/N ← C): 65. Result ID: 699ad13f5698665ec4603bb2.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6ac91e209131cea2a370.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa64a71e209131cea05e65.md>)

[Implementation](<https://github.com/facebookresearch/pythia>)

Recorded evaluation configuration: {"Implementation":"Official"}. This record does not assert independent reproduction.

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