# Multimodal Question Answering on LLaVA-NeXT-7B Evaluation Suite

Source: [Multimodal Question Answering on LLaVA-NeXT-7B Evaluation Suite](<https://www.sota2.com/research/sota/multimodal-question-answering-on-llava-next-7b-evaluation-suite>)

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

Data updated: 2026-07-01T08:41:14.771Z

Evaluation of model performance and efficiency (latency, memory usage, FLOPs) using the LLaVA-NeXT-7B Evaluation Suite for Multimodal Question Answering 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 | Multimodal Question Answering |
| Dataset (including any stated split) | LLaVA-NeXT-7B Evaluation Suite |
| Metric | FLOPs (TFLOPs) |
| Unit | SIZE |
| Direction | Lower is better |
| Results loaded | 6 of 6 |
| 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

- [FLOPs (TFLOPs)](<https://www.sota2.com/research/sota/multimodal-question-answering-on-llava-next-7b-evaluation-suite.md?metric=FLOPs%20(TFLOPs)>): Lower is better; unit: SIZE.

- [E2E Latency (ms)](<https://www.sota2.com/research/sota/multimodal-question-answering-on-llava-next-7b-evaluation-suite.md?metric=E2E%20Latency%20(ms)>): Lower is better; unit: TIME.

- [Prefill Latency (ms)](<https://www.sota2.com/research/sota/multimodal-question-answering-on-llava-next-7b-evaluation-suite.md?metric=Prefill%20Latency%20(ms)>): Lower is better; unit: TIME.

- [Decode Latency / Token (ms)](<https://www.sota2.com/research/sota/multimodal-question-answering-on-llava-next-7b-evaluation-suite.md?metric=Decode%20Latency%20%2F%20Token%20(ms)>): Lower is better; unit: TIME.

- [KV Cache Size (GB)](<https://www.sota2.com/research/sota/multimodal-question-answering-on-llava-next-7b-evaluation-suite.md?metric=KV%20Cache%20Size%20(GB)>): Lower is better; unit: SIZE.

- [GPU Memory Usage (GB)](<https://www.sota2.com/research/sota/multimodal-question-answering-on-llava-next-7b-evaluation-suite.md?metric=GPU%20Memory%20Usage%20(GB)>): Lower is better; unit: SIZE.

- [Average Performance (%)](<https://www.sota2.com/research/sota/multimodal-question-answering-on-llava-next-7b-evaluation-suite.md?metric=Average%20Performance%20(%25)>): Higher is better; unit: PERCENT.

## Recorded results

| Method | FLOPs (TFLOPs) | Result date | Report paper |
| --- | --- | --- | --- |
| VisionZip | 5.6 | Not recorded | 6a449f0ae0224bd4f499303d |
| CDPruner | 5.6 | Not recorded | 6a449f0ae0224bd4f499303d |
| ERA | 5.6 | Not recorded | 6a449f0ae0224bd4f499303d |
| DART | 8.6 | Not recorded | 6a449f0ae0224bd4f499303d |
| PDrop | 17.4 | Not recorded | 6a449f0ae0224bd4f499303d |
| LLaVA-NeXT-7B | 66.1 | Not recorded | 6a449f0ae0224bd4f499303d |

### VisionZip

FLOPs (TFLOPs): 5.6. Result ID: 6a44d2a973d097d845bf9f86.

[Reporting paper](<https://www.sota2.com/research/paper/6a449f0ae0224bd4f499303d.md>)

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

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

Recorded evaluation configuration: {"Token (#)":"160"}. This record does not assert independent reproduction.

### CDPruner

FLOPs (TFLOPs): 5.6. Result ID: 6a44d2a973d097d845bf9f88.

[Reporting paper](<https://www.sota2.com/research/paper/6a449f0ae0224bd4f499303d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79cc1e209131cea7f6cb.md>)

Recorded evaluation configuration: {"Token (#)":"160"}. This record does not assert independent reproduction.

### ERA

FLOPs (TFLOPs): 5.6. Result ID: 6a44d2a973d097d845bf9f89.

[Reporting paper](<https://www.sota2.com/research/paper/6a449f0ae0224bd4f499303d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a449f0ae0224bd4f499303d.md>)

Recorded evaluation configuration: {"Token (#)":"160"}. This record does not assert independent reproduction.

### DART

FLOPs (TFLOPs): 8.6. Result ID: 6a44d2a973d097d845bf9f87.

[Reporting paper](<https://www.sota2.com/research/paper/6a449f0ae0224bd4f499303d.md>)

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

Recorded evaluation configuration: {"Token (#)":"160"}. This record does not assert independent reproduction.

### PDrop

FLOPs (TFLOPs): 17.4. Result ID: 6a44d2a973d097d845bf9f85.

[Reporting paper](<https://www.sota2.com/research/paper/6a449f0ae0224bd4f499303d.md>)

Recorded evaluation configuration: {"Token (#)":"160"}. This record does not assert independent reproduction.

### LLaVA-NeXT-7B

FLOPs (TFLOPs): 66.1. Result ID: 6a44d2a973d097d845bf9f84.

[Reporting paper](<https://www.sota2.com/research/paper/6a449f0ae0224bd4f499303d.md>)

Recorded evaluation configuration: {"Token (#)":"2880"}. This record does not assert independent reproduction.

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