# Vision-Language Understanding on Vision-Language Benchmark Suite Aggregate

Source: [Vision-Language Understanding on Vision-Language Benchmark Suite Aggregate](<https://www.sota2.com/research/sota/vision-language-understanding-on-vision-language-benchmark-suite-aggregate>)

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

Data updated: 2026-04-21T09:45:26.686Z

Evaluation of Vision-Language Understanding models using an aggregate performance metric relative to an established upper bound across various underlying tasks.

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 | Vision-Language Understanding |
| Dataset (including any stated split) | Vision-Language Benchmark Suite Aggregate |
| Metric | Aggregate Performance (%) |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 34 of 34 |
| 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

- [Aggregate Performance (%)](<https://www.sota2.com/research/sota/vision-language-understanding-on-vision-language-benchmark-suite-aggregate.md?metric=Aggregate%20Performance%20(%25)>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Aggregate Performance (%) | Result date | Report paper |
| --- | --- | --- | --- |
| Vanilla | 100 | Not recorded | 69e704c47d0172140ec7ef18 |
| EvoComp | 99.3 | Not recorded | 69e704c47d0172140ec7ef18 |
| VisionZip | 98.8 | Not recorded | 69e704c47d0172140ec7ef18 |
| EvoComp | 98.7 | Not recorded | 69e704c47d0172140ec7ef18 |
| DART | 98.3 | Not recorded | 69e704c47d0172140ec7ef18 |
| EvoComp | 98 | Not recorded | 69e704c47d0172140ec7ef18 |
| EvoComp | 97.8 | Not recorded | 69e704c47d0172140ec7ef18 |
| DART | 97.5 | Not recorded | 69e704c47d0172140ec7ef18 |
| SparseVLM | 97.4 | Not recorded | 69e704c47d0172140ec7ef18 |
| VisionZip | 97.4 | Not recorded | 69e704c47d0172140ec7ef18 |
| PyramidDrop | 97 | Not recorded | 69e704c47d0172140ec7ef18 |
| MustDrop | 97 | Not recorded | 69e704c47d0172140ec7ef18 |
| FastV | 96.9 | Not recorded | 69e704c47d0172140ec7ef18 |
| HiRED | 96.3 | Not recorded | 69e704c47d0172140ec7ef18 |
| SparseVLM | 96.3 | Not recorded | 69e704c47d0172140ec7ef18 |
| PyramidDrop | 95.9 | Not recorded | 69e704c47d0172140ec7ef18 |
| MustDrop | 95.4 | Not recorded | 69e704c47d0172140ec7ef18 |
| EvoComp | 94.9 | Not recorded | 69e704c47d0172140ec7ef18 |
| HiRED | 94.8 | Not recorded | 69e704c47d0172140ec7ef18 |
| FastV | 94.4 | Not recorded | 69e704c47d0172140ec7ef18 |
| DART | 94.2 | Not recorded | 69e704c47d0172140ec7ef18 |
| VisionZip | 94.2 | Not recorded | 69e704c47d0172140ec7ef18 |
| EvoComp | 93.9 | Not recorded | 69e704c47d0172140ec7ef18 |
| DART | 92.8 | Not recorded | 69e704c47d0172140ec7ef18 |
| DART | 91.1 | Not recorded | 69e704c47d0172140ec7ef18 |
| HiRED | 90.8 | Not recorded | 69e704c47d0172140ec7ef18 |
| PruMerge | 90.7 | Not recorded | 69e704c47d0172140ec7ef18 |
| MustDrop | 90 | Not recorded | 69e704c47d0172140ec7ef18 |
| SparseVLM | 89.4 | Not recorded | 69e704c47d0172140ec7ef18 |
| PruMerge | 88.6 | Not recorded | 69e704c47d0172140ec7ef18 |
| FastV | 87.6 | Not recorded | 69e704c47d0172140ec7ef18 |
| PruMerge | 86 | Not recorded | 69e704c47d0172140ec7ef18 |
| PyramidDrop | 84.9 | Not recorded | 69e704c47d0172140ec7ef18 |
| DART | 84.5 | Not recorded | 69e704c47d0172140ec7ef18 |

### Vanilla

Aggregate Performance (%): 100. Result ID: 69e746dc47f34ce924cdeb04.

[Reporting paper](<https://www.sota2.com/research/paper/69e704c47d0172140ec7ef18.md>)

Recorded evaluation configuration: {"Backbone":"LLaVA-1.5-7B","Retained Tokens":"576","Compression Ratio":"100%"}. This record does not assert independent reproduction.

### EvoComp

Aggregate Performance (%): 99.3. Result ID: 69e746dc47f34ce924cdeb0f.

[Reporting paper](<https://www.sota2.com/research/paper/69e704c47d0172140ec7ef18.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69e704c47d0172140ec7ef18.md>)

Recorded evaluation configuration: {"Backbone":"LLaVA-1.5-7B","Retained Tokens":"192","Compression Ratio":"66.7%","Number of compressor layers (l)":"2"}. This record does not assert independent reproduction.

### VisionZip

Aggregate Performance (%): 98.8. Result ID: 69e746dc47f34ce924cdeb08.

[Reporting paper](<https://www.sota2.com/research/paper/69e704c47d0172140ec7ef18.md>)

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

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

Recorded evaluation configuration: {"Backbone":"LLaVA-1.5-7B","Retained Tokens":"192","Compression Ratio":"66.7%"}. This record does not assert independent reproduction.

### EvoComp

Aggregate Performance (%): 98.7. Result ID: 69e746dc47f34ce924cdeb0e.

[Reporting paper](<https://www.sota2.com/research/paper/69e704c47d0172140ec7ef18.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69e704c47d0172140ec7ef18.md>)

Recorded evaluation configuration: {"Backbone":"LLaVA-1.5-7B","Retained Tokens":"192","Compression Ratio":"66.7%","Number of compressor layers (l)":"0"}. This record does not assert independent reproduction.

### DART

Aggregate Performance (%): 98.3. Result ID: 69e746dc47f34ce924cdeb0c.

[Reporting paper](<https://www.sota2.com/research/paper/69e704c47d0172140ec7ef18.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79461e209131cea7cfd7.md>)

Recorded evaluation configuration: {"Backbone":"LLaVA-1.5-7B","Retained Tokens":"192","Compression Ratio":"66.7%"}. This record does not assert independent reproduction.

### EvoComp

Aggregate Performance (%): 98. Result ID: 69e746dc47f34ce924cdeb1a.

[Reporting paper](<https://www.sota2.com/research/paper/69e704c47d0172140ec7ef18.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69e704c47d0172140ec7ef18.md>)

Recorded evaluation configuration: {"Backbone":"LLaVA-1.5-7B","Retained Tokens":"128","Compression Ratio":"77.8%","Number of compressor layers (l)":"2"}. This record does not assert independent reproduction.

### EvoComp

Aggregate Performance (%): 97.8. Result ID: 69e746dc47f34ce924cdeb19.

[Reporting paper](<https://www.sota2.com/research/paper/69e704c47d0172140ec7ef18.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69e704c47d0172140ec7ef18.md>)

Recorded evaluation configuration: {"Backbone":"LLaVA-1.5-7B","Retained Tokens":"128","Compression Ratio":"77.8%","Number of compressor layers (l)":"0"}. This record does not assert independent reproduction.

### DART

Aggregate Performance (%): 97.5. Result ID: 69e746dc47f34ce924cdeb17.

[Reporting paper](<https://www.sota2.com/research/paper/69e704c47d0172140ec7ef18.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79461e209131cea7cfd7.md>)

Recorded evaluation configuration: {"Backbone":"LLaVA-1.5-7B","Retained Tokens":"128","Compression Ratio":"77.8%"}. This record does not assert independent reproduction.

### SparseVLM

Aggregate Performance (%): 97.4. Result ID: 69e746dc47f34ce924cdeb09.

[Reporting paper](<https://www.sota2.com/research/paper/69e704c47d0172140ec7ef18.md>)

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

Recorded evaluation configuration: {"Backbone":"LLaVA-1.5-7B","Retained Tokens":"192","Compression Ratio":"66.7%"}. This record does not assert independent reproduction.

### VisionZip

Aggregate Performance (%): 97.4. Result ID: 69e746dc47f34ce924cdeb13.

[Reporting paper](<https://www.sota2.com/research/paper/69e704c47d0172140ec7ef18.md>)

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

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

Recorded evaluation configuration: {"Backbone":"LLaVA-1.5-7B","Retained Tokens":"128","Compression Ratio":"77.8%"}. This record does not assert independent reproduction.

### PyramidDrop

Aggregate Performance (%): 97. Result ID: 69e746dc47f34ce924cdeb0a.

[Reporting paper](<https://www.sota2.com/research/paper/69e704c47d0172140ec7ef18.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa772c1e209131cea71934.md>)

Recorded evaluation configuration: {"Backbone":"LLaVA-1.5-7B","Retained Tokens":"192","Compression Ratio":"66.7%"}. This record does not assert independent reproduction.

### MustDrop

Aggregate Performance (%): 97. Result ID: 69e746dc47f34ce924cdeb0b.

[Reporting paper](<https://www.sota2.com/research/paper/69e704c47d0172140ec7ef18.md>)

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

Recorded evaluation configuration: {"Backbone":"LLaVA-1.5-7B","Retained Tokens":"192","Compression Ratio":"66.7%"}. This record does not assert independent reproduction.

### FastV

Aggregate Performance (%): 96.9. Result ID: 69e746dc47f34ce924cdeb05.

[Reporting paper](<https://www.sota2.com/research/paper/69e704c47d0172140ec7ef18.md>)

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

Recorded evaluation configuration: {"Backbone":"LLaVA-1.5-7B","Retained Tokens":"192","Compression Ratio":"66.7%"}. This record does not assert independent reproduction.

### HiRED

Aggregate Performance (%): 96.3. Result ID: 69e746dc47f34ce924cdeb06.

[Reporting paper](<https://www.sota2.com/research/paper/69e704c47d0172140ec7ef18.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa76581e209131cea6c829.md>)

Recorded evaluation configuration: {"Backbone":"LLaVA-1.5-7B","Retained Tokens":"192","Compression Ratio":"66.7%"}. This record does not assert independent reproduction.

### SparseVLM

Aggregate Performance (%): 96.3. Result ID: 69e746dc47f34ce924cdeb14.

[Reporting paper](<https://www.sota2.com/research/paper/69e704c47d0172140ec7ef18.md>)

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

Recorded evaluation configuration: {"Backbone":"LLaVA-1.5-7B","Retained Tokens":"128","Compression Ratio":"77.8%"}. This record does not assert independent reproduction.

### PyramidDrop

Aggregate Performance (%): 95.9. Result ID: 69e746dc47f34ce924cdeb15.

[Reporting paper](<https://www.sota2.com/research/paper/69e704c47d0172140ec7ef18.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa772c1e209131cea71934.md>)

Recorded evaluation configuration: {"Backbone":"LLaVA-1.5-7B","Retained Tokens":"128","Compression Ratio":"77.8%"}. This record does not assert independent reproduction.

### MustDrop

Aggregate Performance (%): 95.4. Result ID: 69e746dc47f34ce924cdeb16.

[Reporting paper](<https://www.sota2.com/research/paper/69e704c47d0172140ec7ef18.md>)

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

Recorded evaluation configuration: {"Backbone":"LLaVA-1.5-7B","Retained Tokens":"128","Compression Ratio":"77.8%"}. This record does not assert independent reproduction.

### EvoComp

Aggregate Performance (%): 94.9. Result ID: 69e746dc47f34ce924cdeb24.

[Reporting paper](<https://www.sota2.com/research/paper/69e704c47d0172140ec7ef18.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69e704c47d0172140ec7ef18.md>)

Recorded evaluation configuration: {"Backbone":"LLaVA-1.5-7B","Retained Tokens":"64","Compression Ratio":"88.9%","Number of compressor layers (l)":"0"}. This record does not assert independent reproduction.

### HiRED

Aggregate Performance (%): 94.8. Result ID: 69e746dc47f34ce924cdeb11.

[Reporting paper](<https://www.sota2.com/research/paper/69e704c47d0172140ec7ef18.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa76581e209131cea6c829.md>)

Recorded evaluation configuration: {"Backbone":"LLaVA-1.5-7B","Retained Tokens":"128","Compression Ratio":"77.8%"}. This record does not assert independent reproduction.

### FastV

Aggregate Performance (%): 94.4. Result ID: 69e746dc47f34ce924cdeb10.

[Reporting paper](<https://www.sota2.com/research/paper/69e704c47d0172140ec7ef18.md>)

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

Recorded evaluation configuration: {"Backbone":"LLaVA-1.5-7B","Retained Tokens":"128","Compression Ratio":"77.8%"}. This record does not assert independent reproduction.

### DART

Aggregate Performance (%): 94.2. Result ID: 69e746dc47f34ce924cdeb0d.

[Reporting paper](<https://www.sota2.com/research/paper/69e704c47d0172140ec7ef18.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79461e209131cea7cfd7.md>)

Recorded evaluation configuration: {"Backbone":"LLaVA-1.5-7B","Retained Tokens":"192","Compression Ratio":"66.7%","Number of compressor layers (l)":"0"}. This record does not assert independent reproduction.

### VisionZip

Aggregate Performance (%): 94.2. Result ID: 69e746dc47f34ce924cdeb1e.

[Reporting paper](<https://www.sota2.com/research/paper/69e704c47d0172140ec7ef18.md>)

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

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

Recorded evaluation configuration: {"Backbone":"LLaVA-1.5-7B","Retained Tokens":"64","Compression Ratio":"88.9%"}. This record does not assert independent reproduction.

### EvoComp

Aggregate Performance (%): 93.9. Result ID: 69e746dc47f34ce924cdeb25.

[Reporting paper](<https://www.sota2.com/research/paper/69e704c47d0172140ec7ef18.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69e704c47d0172140ec7ef18.md>)

Recorded evaluation configuration: {"Backbone":"LLaVA-1.5-7B","Retained Tokens":"64","Compression Ratio":"88.9%","Number of compressor layers (l)":"2"}. This record does not assert independent reproduction.

### DART

Aggregate Performance (%): 92.8. Result ID: 69e746dc47f34ce924cdeb22.

[Reporting paper](<https://www.sota2.com/research/paper/69e704c47d0172140ec7ef18.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79461e209131cea7cfd7.md>)

Recorded evaluation configuration: {"Backbone":"LLaVA-1.5-7B","Retained Tokens":"64","Compression Ratio":"88.9%"}. This record does not assert independent reproduction.

### DART

Aggregate Performance (%): 91.1. Result ID: 69e746dc47f34ce924cdeb18.

[Reporting paper](<https://www.sota2.com/research/paper/69e704c47d0172140ec7ef18.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79461e209131cea7cfd7.md>)

Recorded evaluation configuration: {"Backbone":"LLaVA-1.5-7B","Retained Tokens":"128","Compression Ratio":"77.8%","Number of compressor layers (l)":"0"}. This record does not assert independent reproduction.

### HiRED

Aggregate Performance (%): 90.8. Result ID: 69e746dc47f34ce924cdeb1c.

[Reporting paper](<https://www.sota2.com/research/paper/69e704c47d0172140ec7ef18.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa76581e209131cea6c829.md>)

Recorded evaluation configuration: {"Backbone":"LLaVA-1.5-7B","Retained Tokens":"64","Compression Ratio":"88.9%"}. This record does not assert independent reproduction.

### PruMerge

Aggregate Performance (%): 90.7. Result ID: 69e746dc47f34ce924cdeb07.

[Reporting paper](<https://www.sota2.com/research/paper/69e704c47d0172140ec7ef18.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73411e209131cea5b51f.md>)

Recorded evaluation configuration: {"Backbone":"LLaVA-1.5-7B","Retained Tokens":"192","Compression Ratio":"66.7%"}. This record does not assert independent reproduction.

### MustDrop

Aggregate Performance (%): 90. Result ID: 69e746dc47f34ce924cdeb21.

[Reporting paper](<https://www.sota2.com/research/paper/69e704c47d0172140ec7ef18.md>)

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

Recorded evaluation configuration: {"Backbone":"LLaVA-1.5-7B","Retained Tokens":"64","Compression Ratio":"88.9%"}. This record does not assert independent reproduction.

### SparseVLM

Aggregate Performance (%): 89.4. Result ID: 69e746dc47f34ce924cdeb1f.

[Reporting paper](<https://www.sota2.com/research/paper/69e704c47d0172140ec7ef18.md>)

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

Recorded evaluation configuration: {"Backbone":"LLaVA-1.5-7B","Retained Tokens":"64","Compression Ratio":"88.9%"}. This record does not assert independent reproduction.

### PruMerge

Aggregate Performance (%): 88.6. Result ID: 69e746dc47f34ce924cdeb12.

[Reporting paper](<https://www.sota2.com/research/paper/69e704c47d0172140ec7ef18.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73411e209131cea5b51f.md>)

Recorded evaluation configuration: {"Backbone":"LLaVA-1.5-7B","Retained Tokens":"128","Compression Ratio":"77.8%"}. This record does not assert independent reproduction.

### FastV

Aggregate Performance (%): 87.6. Result ID: 69e746dc47f34ce924cdeb1b.

[Reporting paper](<https://www.sota2.com/research/paper/69e704c47d0172140ec7ef18.md>)

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

Recorded evaluation configuration: {"Backbone":"LLaVA-1.5-7B","Retained Tokens":"64","Compression Ratio":"88.9%"}. This record does not assert independent reproduction.

### PruMerge

Aggregate Performance (%): 86. Result ID: 69e746dc47f34ce924cdeb1d.

[Reporting paper](<https://www.sota2.com/research/paper/69e704c47d0172140ec7ef18.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73411e209131cea5b51f.md>)

Recorded evaluation configuration: {"Backbone":"LLaVA-1.5-7B","Retained Tokens":"64","Compression Ratio":"88.9%"}. This record does not assert independent reproduction.

### PyramidDrop

Aggregate Performance (%): 84.9. Result ID: 69e746dc47f34ce924cdeb20.

[Reporting paper](<https://www.sota2.com/research/paper/69e704c47d0172140ec7ef18.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa772c1e209131cea71934.md>)

Recorded evaluation configuration: {"Backbone":"LLaVA-1.5-7B","Retained Tokens":"64","Compression Ratio":"88.9%"}. This record does not assert independent reproduction.

### DART

Aggregate Performance (%): 84.5. Result ID: 69e746dc47f34ce924cdeb23.

[Reporting paper](<https://www.sota2.com/research/paper/69e704c47d0172140ec7ef18.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79461e209131cea7cfd7.md>)

Recorded evaluation configuration: {"Backbone":"LLaVA-1.5-7B","Retained Tokens":"64","Compression Ratio":"88.9%","Number of compressor layers (l)":"0"}. This record does not assert independent reproduction.

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