# Multi-turn Visual Question Answering on MT-GQA

Source: [Multi-turn Visual Question Answering on MT-GQA](<https://www.sota2.com/research/sota/multi-turn-visual-question-answering-on-mt-gqa>)

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

Data updated: 2026-03-24T16:03:39.695Z

Evaluation of model performance on the MT-GQA dataset for multi-turn visual question answering, focusing on latency (TTFT, E2ET) and computational efficiency (Memory, TFLOPs).

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 | Multi-turn Visual Question Answering |
| Dataset (including any stated split) | MT-GQA |
| Metric | Time To First Token (TTFT) |
| Unit | TIME |
| Direction | Lower is better |
| Results loaded | 11 of 11 |
| 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

- [Time To First Token (TTFT)](<https://www.sota2.com/research/sota/multi-turn-visual-question-answering-on-mt-gqa.md?metric=Time%20To%20First%20Token%20(TTFT)>): Lower is better; unit: TIME.

- [End-to-End Time (E2ET)](<https://www.sota2.com/research/sota/multi-turn-visual-question-answering-on-mt-gqa.md?metric=End-to-End%20Time%20(E2ET)>): Lower is better; unit: TIME.

- [Memory Usage (Mem.)](<https://www.sota2.com/research/sota/multi-turn-visual-question-answering-on-mt-gqa.md?metric=Memory%20Usage%20(Mem.)>): Lower is better; unit: SIZE.

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

## Recorded results

| Method | Time To First Token (TTFT) | Result date | Report paper |
| --- | --- | --- | --- |
| Sample | 96.9 | Not recorded | 69c21aa60636e5bf4afc4a72 |
| MetaCompress | 97.8 | Not recorded | 69c21aa60636e5bf4afc4a72 |
| Random | 98.2 | Not recorded | 69c21aa60636e5bf4afc4a72 |
| FastV | 102 | Not recorded | 69c21aa60636e5bf4afc4a72 |
| PruMerge | 107 | Not recorded | 69c21aa60636e5bf4afc4a72 |
| Random | 174 | Not recorded | 69c21aa60636e5bf4afc4a72 |
| MetaCompress | 174 | Not recorded | 69c21aa60636e5bf4afc4a72 |
| Sample | 176 | Not recorded | 69c21aa60636e5bf4afc4a72 |
| FastV | 219 | Not recorded | 69c21aa60636e5bf4afc4a72 |
| Base | 232 | Not recorded | 69c21aa60636e5bf4afc4a72 |
| Base | 484 | Not recorded | 69c21aa60636e5bf4afc4a72 |

### Sample

Time To First Token (TTFT): 96.9. Result ID: 69c2b5c847f34ce924ac0112.

[Reporting paper](<https://www.sota2.com/research/paper/69c21aa60636e5bf4afc4a72.md>)

Recorded evaluation configuration: {"Model":"LLaVA-1.5-7b","Reduction Rate":"90%"}. This record does not assert independent reproduction.

### MetaCompress

Time To First Token (TTFT): 97.8. Result ID: 69c2b5c847f34ce924ac0115.

[Reporting paper](<https://www.sota2.com/research/paper/69c21aa60636e5bf4afc4a72.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69c21aa60636e5bf4afc4a72.md>)

Recorded evaluation configuration: {"Model":"LLaVA-1.5-7b","Reduction Rate":"90%"}. This record does not assert independent reproduction.

### Random

Time To First Token (TTFT): 98.2. Result ID: 69c2b5c847f34ce924ac0111.

[Reporting paper](<https://www.sota2.com/research/paper/69c21aa60636e5bf4afc4a72.md>)

Recorded evaluation configuration: {"Model":"LLaVA-1.5-7b","Reduction Rate":"90%"}. This record does not assert independent reproduction.

### FastV

Time To First Token (TTFT): 102. Result ID: 69c2b5c847f34ce924ac0113.

[Reporting paper](<https://www.sota2.com/research/paper/69c21aa60636e5bf4afc4a72.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6a0b1e209131cea25b04.md>)

[Implementation](<https://github.com/google-research/vision_transformer>)

Recorded evaluation configuration: {"Model":"LLaVA-1.5-7b","Reduction Rate":"90%"}. This record does not assert independent reproduction.

### PruMerge

Time To First Token (TTFT): 107. Result ID: 69c2b5c847f34ce924ac0114.

[Reporting paper](<https://www.sota2.com/research/paper/69c21aa60636e5bf4afc4a72.md>)

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

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

Recorded evaluation configuration: {"Model":"LLaVA-1.5-7b","Reduction Rate":"90%"}. This record does not assert independent reproduction.

### Random

Time To First Token (TTFT): 174. Result ID: 69c2b5c847f34ce924ac0117.

[Reporting paper](<https://www.sota2.com/research/paper/69c21aa60636e5bf4afc4a72.md>)

Recorded evaluation configuration: {"Model":"LLaVA-NeXT-7b","Reduction Rate":"90%"}. This record does not assert independent reproduction.

### MetaCompress

Time To First Token (TTFT): 174. Result ID: 69c2b5c847f34ce924ac011a.

[Reporting paper](<https://www.sota2.com/research/paper/69c21aa60636e5bf4afc4a72.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69c21aa60636e5bf4afc4a72.md>)

Recorded evaluation configuration: {"Model":"LLaVA-NeXT-7b","Reduction Rate":"90%"}. This record does not assert independent reproduction.

### Sample

Time To First Token (TTFT): 176. Result ID: 69c2b5c847f34ce924ac0118.

[Reporting paper](<https://www.sota2.com/research/paper/69c21aa60636e5bf4afc4a72.md>)

Recorded evaluation configuration: {"Model":"LLaVA-NeXT-7b","Reduction Rate":"90%"}. This record does not assert independent reproduction.

### FastV

Time To First Token (TTFT): 219. Result ID: 69c2b5c847f34ce924ac0119.

[Reporting paper](<https://www.sota2.com/research/paper/69c21aa60636e5bf4afc4a72.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6a0b1e209131cea25b04.md>)

[Implementation](<https://github.com/google-research/vision_transformer>)

Recorded evaluation configuration: {"Model":"LLaVA-NeXT-7b","Reduction Rate":"90%"}. This record does not assert independent reproduction.

### Base

Time To First Token (TTFT): 232. Result ID: 69c2b5c847f34ce924ac0110.

[Reporting paper](<https://www.sota2.com/research/paper/69c21aa60636e5bf4afc4a72.md>)

Recorded evaluation configuration: {"Model":"LLaVA-1.5-7b","Reduction Rate":"0%"}. This record does not assert independent reproduction.

### Base

Time To First Token (TTFT): 484. Result ID: 69c2b5c847f34ce924ac0116.

[Reporting paper](<https://www.sota2.com/research/paper/69c21aa60636e5bf4afc4a72.md>)

Recorded evaluation configuration: {"Model":"LLaVA-NeXT-7b","Reduction Rate":"0%"}. This record does not assert independent reproduction.

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