# Efficiency Metrics for Visual Question Answering on GQA

Source: [Efficiency Metrics for Visual Question Answering on GQA](<https://www.sota2.com/research/sota/efficiency-metrics-for-visual-question-answering-on-gqa>)

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

Data updated: 2026-02-26T06:06:33.634Z

Evaluation of model efficiency, latency, and hardware utilization (TFLOPs, Memory, Throughput) for Visual Question Answering tasks on the GQA dataset.

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) | GQA |
| Metric | Token Count |
| Unit | SCALAR |
| Direction | Higher 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

- [Token Count](<https://www.sota2.com/research/sota/efficiency-metrics-for-visual-question-answering-on-gqa.md?metric=Token%20Count>): Higher is better; unit: SCALAR.

- [TFLOPs](<https://www.sota2.com/research/sota/efficiency-metrics-for-visual-question-answering-on-gqa.md?metric=TFLOPs>): Lower is better; unit: SCALAR.

- [GPU Memory (GB)](<https://www.sota2.com/research/sota/efficiency-metrics-for-visual-question-answering-on-gqa.md?metric=GPU%20Memory%20(GB)>): Lower is better; unit: SIZE.

- [Response Time (s)](<https://www.sota2.com/research/sota/efficiency-metrics-for-visual-question-answering-on-gqa.md?metric=Response%20Time%20(s)>): Lower is better; unit: TIME.

- [Throughput (samples/s)](<https://www.sota2.com/research/sota/efficiency-metrics-for-visual-question-answering-on-gqa.md?metric=Throughput%20(samples%2Fs)>): Higher is better; unit: SCALAR.

## Recorded results

| Method | Token Count | Result date | Report paper |
| --- | --- | --- | --- |
| InternVL2 | 1699 | Not recorded | 68fa760f1e209131cea6b31c |
| IMP | 729 | Not recorded | 68fa760f1e209131cea6b31c |
| LLaVA | 576 | Not recorded | 68fa760f1e209131cea6b31c |
| FlashSloth-HD | 414 | Not recorded | 68fa760f1e209131cea6b31c |
| Qwen2-VL | 352 | Not recorded | 68fa760f1e209131cea6b31c |
| FlashSloth | 90 | Not recorded | 68fa760f1e209131cea6b31c |

### InternVL2

Token Count: 1699. Result ID: 69940cdf3a6c3148491ab61f.

[Reporting paper](<https://www.sota2.com/research/paper/68fa760f1e209131cea6b31c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73081e209131cea59f8b.md>)

[Implementation](<https://github.com/OpenGVLab/InternVL>)

Recorded evaluation configuration: {"Parameters":"2B"}. This record does not assert independent reproduction.

### IMP

Token Count: 729. Result ID: 69940cdf3a6c3148491ab61d.

[Reporting paper](<https://www.sota2.com/research/paper/68fa760f1e209131cea6b31c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa736c1e209131cea5c6ab.md>)

[Implementation](<https://github.com/MILVLG/imp>)

Recorded evaluation configuration: {"Parameters":"3.1B"}. This record does not assert independent reproduction.

### LLaVA

Token Count: 576. Result ID: 69940cdf3a6c3148491ab61c.

[Reporting paper](<https://www.sota2.com/research/paper/68fa760f1e209131cea6b31c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa732e1e209131cea5af59.md>)

[Implementation](<https://llava-vl.github.io/>)

Recorded evaluation configuration: {"Parameters":"7B","Version":"V1.5"}. This record does not assert independent reproduction.

### FlashSloth-HD

Token Count: 414. Result ID: 69940cdf3a6c3148491ab620.

[Reporting paper](<https://www.sota2.com/research/paper/68fa760f1e209131cea6b31c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa760f1e209131cea6b31c.md>)

[Implementation](<https://github.com/codefanw/FlashSloth>)

Recorded evaluation configuration: {"Parameters":"3.2B"}. This record does not assert independent reproduction.

### Qwen2-VL

Token Count: 352. Result ID: 69940cdf3a6c3148491ab61e.

[Reporting paper](<https://www.sota2.com/research/paper/68fa760f1e209131cea6b31c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75151e209131cea65976.md>)

[Implementation](<https://github.com/QwenLM/Qwen2-VL>)

Recorded evaluation configuration: {"Parameters":"2B"}. This record does not assert independent reproduction.

### FlashSloth

Token Count: 90. Result ID: 69940cdf3a6c3148491ab621.

[Reporting paper](<https://www.sota2.com/research/paper/68fa760f1e209131cea6b31c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa760f1e209131cea6b31c.md>)

[Implementation](<https://github.com/codefanw/FlashSloth>)

Recorded evaluation configuration: {"Parameters":"3.2B"}. This record does not assert independent reproduction.

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