# 3D/4D Visual Question Answering on MMSI Bench 1.0 (test)

Source: [3D/4D Visual Question Answering on MMSI Bench 1.0 (test)](<https://www.sota2.com/research/sota/3d-4d-visual-question-answering-on-mmsi-bench-1-0-test>)

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

Data updated: 2026-02-26T05:36:14.972Z

Evaluation of 3D/4D Visual Question Answering performance on the test split of the MMSI Bench 1.0 dataset, reporting average multiple choice 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 | 3D/4D Visual Question Answering |
| Dataset (including any stated split) | MMSI Bench 1.0 (test) |
| Metric | Avg Multiple Choice Accuracy |
| 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

- [Avg Multiple Choice Accuracy](<https://www.sota2.com/research/sota/3d-4d-visual-question-answering-on-mmsi-bench-1-0-test.md?metric=Avg%20Multiple%20Choice%20Accuracy>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Avg Multiple Choice Accuracy | Result date | Report paper |
| --- | --- | --- | --- |
| 4D-RGPT-8B | 33.3 | Not recorded | 6948d086423ea42265729f78 |
| 4D-RGPT-15B | 31.7 | Not recorded | 6948d086423ea42265729f78 |
| NVILA-Lite-8B | 31.3 | Not recorded | 6948d086423ea42265729f78 |
| NVILA-Lite-15B | 29.5 | Not recorded | 6948d086423ea42265729f78 |

### 4D-RGPT-8B

Avg Multiple Choice Accuracy: 33.3. Result ID: 69982c705698665ec487662f.

[Reporting paper](<https://www.sota2.com/research/paper/6948d086423ea42265729f78.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6948d086423ea42265729f78.md>)

Recorded evaluation configuration: {"parameter\_scale":"8B"}. This record does not assert independent reproduction.

### 4D-RGPT-15B

Avg Multiple Choice Accuracy: 31.7. Result ID: 69982c705698665ec4876631.

[Reporting paper](<https://www.sota2.com/research/paper/6948d086423ea42265729f78.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6948d086423ea42265729f78.md>)

Recorded evaluation configuration: {"parameter\_scale":"15B"}. This record does not assert independent reproduction.

### NVILA-Lite-8B

Avg Multiple Choice Accuracy: 31.3. Result ID: 69982c705698665ec487662e.

[Reporting paper](<https://www.sota2.com/research/paper/6948d086423ea42265729f78.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa77481e209131cea723b2.md>)

Recorded evaluation configuration: {"parameter\_scale":"8B"}. This record does not assert independent reproduction.

### NVILA-Lite-15B

Avg Multiple Choice Accuracy: 29.5. Result ID: 69982c705698665ec4876630.

[Reporting paper](<https://www.sota2.com/research/paper/6948d086423ea42265729f78.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa77481e209131cea723b2.md>)

Recorded evaluation configuration: {"parameter\_scale":"15B"}. This record does not assert independent reproduction.

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