# Multimodal Evaluation on Combined Benchmark Suite (GQA, MMB, MME, VQA-T, SQA-I, VQA-v2)

Source: [Multimodal Evaluation on Combined Benchmark Suite (GQA, MMB, MME, VQA-T, SQA-I, VQA-v2)](<https://www.sota2.com/research/sota/multimodal-evaluation-on-combined-benchmark-suite-gqa-mmb-mme-vqa-t-sqa-i-vqa-v2>)

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

Data updated: 2026-04-13T07:29:36.005Z

Evaluation of multimodal capabilities across the combined suite including GQA, MMB, MME, VQA-T, SQA-I, and VQA-v2 datasets.

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 | Multimodal Evaluation |
| Dataset (including any stated split) | Combined Benchmark Suite (GQA, MMB, MME, VQA-T, SQA-I, VQA-v2) |
| Metric | Relative Accuracy |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 28 of 28 |
| 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

- [Relative Accuracy](<https://www.sota2.com/research/sota/multimodal-evaluation-on-combined-benchmark-suite-gqa-mmb-mme-vqa-t-sqa-i-vqa-v2.md?metric=Relative%20Accuracy>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Relative Accuracy | Result date | Report paper |
| --- | --- | --- | --- |
| LLaVA-NeXT-7B | 100 | Not recorded | 68fa7a1c1e209131cea81398 |
| LLaVA-NeXT-7B | 100 | Not recorded | 68fa7a1c1e209131cea81398 |
| TwigVLM | 99 | Not recorded | 68fa7a1c1e209131cea81398 |
| TwigVLM++ | 99 | Not recorded | 68fa7a1c1e209131cea81398 |
| TwigVLM | 99 | Not recorded | 68fa7a1c1e209131cea81398 |
| TwigVLM++ | 99 | Not recorded | 68fa7a1c1e209131cea81398 |
| FastV | 97.6 | Not recorded | 68fa7a1c1e209131cea81398 |
| FastV | 97.6 | Not recorded | 68fa7a1c1e209131cea81398 |
| VisionZip‡ | 97.4 | Not recorded | 68fa7a1c1e209131cea81398 |
| VisionZip‡ | 97.4 | Not recorded | 68fa7a1c1e209131cea81398 |
| VisionZip | 97.1 | Not recorded | 68fa7a1c1e209131cea81398 |
| TwigVLM++ | 97.1 | Not recorded | 68fa7a1c1e209131cea81398 |
| VisionZip | 97.1 | Not recorded | 68fa7a1c1e209131cea81398 |
| TwigVLM++ | 97.1 | Not recorded | 68fa7a1c1e209131cea81398 |
| TwigVLM | 96.2 | Not recorded | 68fa7a1c1e209131cea81398 |
| TwigVLM | 96.2 | Not recorded | 68fa7a1c1e209131cea81398 |
| VisionZip‡ | 95.8 | Not recorded | 68fa7a1c1e209131cea81398 |
| VisionZip‡ | 95.8 | Not recorded | 68fa7a1c1e209131cea81398 |
| SparseVLM | 95.4 | Not recorded | 68fa7a1c1e209131cea81398 |
| SparseVLM | 95.4 | Not recorded | 68fa7a1c1e209131cea81398 |
| VisionZip | 93.8 | Not recorded | 68fa7a1c1e209131cea81398 |
| VisionZip | 93.8 | Not recorded | 68fa7a1c1e209131cea81398 |
| MustDrop | 92.6 | Not recorded | 68fa7a1c1e209131cea81398 |
| MustDrop | 92.6 | Not recorded | 68fa7a1c1e209131cea81398 |
| SparseVLM | 92.3 | Not recorded | 68fa7a1c1e209131cea81398 |
| SparseVLM | 92.3 | Not recorded | 68fa7a1c1e209131cea81398 |
| FastV | 88.2 | Not recorded | 68fa7a1c1e209131cea81398 |
| FastV | 88.2 | Not recorded | 68fa7a1c1e209131cea81398 |

### LLaVA-NeXT-7B

Relative Accuracy: 100. Result ID: 69dc9a8e47f34ce924c339b3.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a1c1e209131cea81398.md>)

Recorded evaluation configuration: {"Retained Tokens":"2880","Base Model":"LLaVA-NeXT-7B"}. This record does not assert independent reproduction.

### LLaVA-NeXT-7B

Relative Accuracy: 100. Result ID: 69dc9a8e47f34ce924c33a3f.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a1c1e209131cea81398.md>)

Recorded evaluation configuration: {"Retained Tokens":"2880","Base Model":"LLaVA-NeXT-7B"}. This record does not assert independent reproduction.

### TwigVLM

Relative Accuracy: 99. Result ID: 69dc9a8e47f34ce924c339b8.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a1c1e209131cea81398.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7a1c1e209131cea81398.md>)

Recorded evaluation configuration: {"Retained Tokens":"640","Base Model":"LLaVA-NeXT-7B"}. This record does not assert independent reproduction.

### TwigVLM++

Relative Accuracy: 99. Result ID: 69dc9a8e47f34ce924c339b9.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a1c1e209131cea81398.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7a1c1e209131cea81398.md>)

Recorded evaluation configuration: {"Retained Tokens":"640","Base Model":"LLaVA-NeXT-7B"}. This record does not assert independent reproduction.

### TwigVLM

Relative Accuracy: 99. Result ID: 69dc9a8e47f34ce924c33a44.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a1c1e209131cea81398.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7a1c1e209131cea81398.md>)

Recorded evaluation configuration: {"Retained Tokens":"640","Base Model":"LLaVA-NeXT-7B"}. This record does not assert independent reproduction.

### TwigVLM++

Relative Accuracy: 99. Result ID: 69dc9a8e47f34ce924c33a45.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a1c1e209131cea81398.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7a1c1e209131cea81398.md>)

Recorded evaluation configuration: {"Retained Tokens":"640","Base Model":"LLaVA-NeXT-7B"}. This record does not assert independent reproduction.

### FastV

Relative Accuracy: 97.6. Result ID: 69dc9a8e47f34ce924c339b4.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a1c1e209131cea81398.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa74ae1e209131cea632b7.md>)

Recorded evaluation configuration: {"Retained Tokens":"640","Base Model":"LLaVA-NeXT-7B"}. This record does not assert independent reproduction.

### FastV

Relative Accuracy: 97.6. Result ID: 69dc9a8e47f34ce924c33a40.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a1c1e209131cea81398.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa74ae1e209131cea632b7.md>)

Recorded evaluation configuration: {"Retained Tokens":"640","Base Model":"LLaVA-NeXT-7B"}. This record does not assert independent reproduction.

### VisionZip‡

Relative Accuracy: 97.4. Result ID: 69dc9a8e47f34ce924c339b7.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a1c1e209131cea81398.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa760f1e209131cea6b33b.md>)

[Implementation](<https://github.com/dvlab-research/VisionZip>)

Recorded evaluation configuration: {"Retained Tokens":"640","Base Model":"LLaVA-NeXT-7B"}. This record does not assert independent reproduction.

### VisionZip‡

Relative Accuracy: 97.4. Result ID: 69dc9a8e47f34ce924c33a43.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a1c1e209131cea81398.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa760f1e209131cea6b33b.md>)

[Implementation](<https://github.com/dvlab-research/VisionZip>)

Recorded evaluation configuration: {"Retained Tokens":"640","Base Model":"LLaVA-NeXT-7B"}. This record does not assert independent reproduction.

### VisionZip

Relative Accuracy: 97.1. Result ID: 69dc9a8e47f34ce924c339b6.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a1c1e209131cea81398.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa760f1e209131cea6b33b.md>)

[Implementation](<https://github.com/dvlab-research/VisionZip>)

Recorded evaluation configuration: {"Retained Tokens":"640","Base Model":"LLaVA-NeXT-7B"}. This record does not assert independent reproduction.

### TwigVLM++

Relative Accuracy: 97.1. Result ID: 69dc9a8e47f34ce924c339c0.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a1c1e209131cea81398.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7a1c1e209131cea81398.md>)

Recorded evaluation configuration: {"Retained Tokens":"320","Base Model":"LLaVA-NeXT-7B"}. This record does not assert independent reproduction.

### VisionZip

Relative Accuracy: 97.1. Result ID: 69dc9a8e47f34ce924c33a42.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a1c1e209131cea81398.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa760f1e209131cea6b33b.md>)

[Implementation](<https://github.com/dvlab-research/VisionZip>)

Recorded evaluation configuration: {"Retained Tokens":"640","Base Model":"LLaVA-NeXT-7B"}. This record does not assert independent reproduction.

### TwigVLM++

Relative Accuracy: 97.1. Result ID: 69dc9a8e47f34ce924c33a4c.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a1c1e209131cea81398.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7a1c1e209131cea81398.md>)

Recorded evaluation configuration: {"Retained Tokens":"320","Base Model":"LLaVA-NeXT-7B"}. This record does not assert independent reproduction.

### TwigVLM

Relative Accuracy: 96.2. Result ID: 69dc9a8e47f34ce924c339bf.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a1c1e209131cea81398.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7a1c1e209131cea81398.md>)

Recorded evaluation configuration: {"Retained Tokens":"320","Base Model":"LLaVA-NeXT-7B"}. This record does not assert independent reproduction.

### TwigVLM

Relative Accuracy: 96.2. Result ID: 69dc9a8e47f34ce924c33a4b.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a1c1e209131cea81398.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7a1c1e209131cea81398.md>)

Recorded evaluation configuration: {"Retained Tokens":"320","Base Model":"LLaVA-NeXT-7B"}. This record does not assert independent reproduction.

### VisionZip‡

Relative Accuracy: 95.8. Result ID: 69dc9a8e47f34ce924c339be.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a1c1e209131cea81398.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa760f1e209131cea6b33b.md>)

[Implementation](<https://github.com/dvlab-research/VisionZip>)

Recorded evaluation configuration: {"Retained Tokens":"320","Base Model":"LLaVA-NeXT-7B"}. This record does not assert independent reproduction.

### VisionZip‡

Relative Accuracy: 95.8. Result ID: 69dc9a8e47f34ce924c33a4a.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a1c1e209131cea81398.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa760f1e209131cea6b33b.md>)

[Implementation](<https://github.com/dvlab-research/VisionZip>)

Recorded evaluation configuration: {"Retained Tokens":"320","Base Model":"LLaVA-NeXT-7B"}. This record does not assert independent reproduction.

### SparseVLM

Relative Accuracy: 95.4. Result ID: 69dc9a8e47f34ce924c339b5.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a1c1e209131cea81398.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79181e209131cea7c1b1.md>)

Recorded evaluation configuration: {"Retained Tokens":"640","Base Model":"LLaVA-NeXT-7B"}. This record does not assert independent reproduction.

### SparseVLM

Relative Accuracy: 95.4. Result ID: 69dc9a8e47f34ce924c33a41.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a1c1e209131cea81398.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79181e209131cea7c1b1.md>)

Recorded evaluation configuration: {"Retained Tokens":"640","Base Model":"LLaVA-NeXT-7B"}. This record does not assert independent reproduction.

### VisionZip

Relative Accuracy: 93.8. Result ID: 69dc9a8e47f34ce924c339bd.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a1c1e209131cea81398.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa760f1e209131cea6b33b.md>)

[Implementation](<https://github.com/dvlab-research/VisionZip>)

Recorded evaluation configuration: {"Retained Tokens":"320","Base Model":"LLaVA-NeXT-7B"}. This record does not assert independent reproduction.

### VisionZip

Relative Accuracy: 93.8. Result ID: 69dc9a8e47f34ce924c33a49.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a1c1e209131cea81398.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa760f1e209131cea6b33b.md>)

[Implementation](<https://github.com/dvlab-research/VisionZip>)

Recorded evaluation configuration: {"Retained Tokens":"320","Base Model":"LLaVA-NeXT-7B"}. This record does not assert independent reproduction.

### MustDrop

Relative Accuracy: 92.6. Result ID: 69dc9a8e47f34ce924c339bc.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a1c1e209131cea81398.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75c41e209131cea69d42.md>)

Recorded evaluation configuration: {"Retained Tokens":"320","Base Model":"LLaVA-NeXT-7B"}. This record does not assert independent reproduction.

### MustDrop

Relative Accuracy: 92.6. Result ID: 69dc9a8e47f34ce924c33a48.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a1c1e209131cea81398.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75c41e209131cea69d42.md>)

Recorded evaluation configuration: {"Retained Tokens":"320","Base Model":"LLaVA-NeXT-7B"}. This record does not assert independent reproduction.

### SparseVLM

Relative Accuracy: 92.3. Result ID: 69dc9a8e47f34ce924c339bb.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a1c1e209131cea81398.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79181e209131cea7c1b1.md>)

Recorded evaluation configuration: {"Retained Tokens":"320","Base Model":"LLaVA-NeXT-7B"}. This record does not assert independent reproduction.

### SparseVLM

Relative Accuracy: 92.3. Result ID: 69dc9a8e47f34ce924c33a47.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a1c1e209131cea81398.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79181e209131cea7c1b1.md>)

Recorded evaluation configuration: {"Retained Tokens":"320","Base Model":"LLaVA-NeXT-7B"}. This record does not assert independent reproduction.

### FastV

Relative Accuracy: 88.2. Result ID: 69dc9a8e47f34ce924c339ba.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a1c1e209131cea81398.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa74ae1e209131cea632b7.md>)

Recorded evaluation configuration: {"Retained Tokens":"320","Base Model":"LLaVA-NeXT-7B"}. This record does not assert independent reproduction.

### FastV

Relative Accuracy: 88.2. Result ID: 69dc9a8e47f34ce924c33a46.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a1c1e209131cea81398.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa74ae1e209131cea632b7.md>)

Recorded evaluation configuration: {"Retained Tokens":"320","Base Model":"LLaVA-NeXT-7B"}. This record does not assert independent reproduction.

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