# Text-to-Video Generation on VBench (Quality and Efficiency Metrics)

Source: [Text-to-Video Generation on VBench (Quality and Efficiency Metrics)](<https://www.sota2.com/research/sota/text-to-video-generation-on-vbench-quality-and-efficiency-metrics>)

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

Data updated: 2026-04-07T08:16:59.822Z

Evaluation of Text-to-Video Generation models on the VBench dataset, covering semantic accuracy, perceptual quality, and operational efficiency metrics like latency and speedup.

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 | Text-to-Video Generation |
| Dataset (including any stated split) | VBench |
| Metric | VBench Semantic Score |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 17 of 17 |
| 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

- [VBench Semantic Score](<https://www.sota2.com/research/sota/text-to-video-generation-on-vbench-quality-and-efficiency-metrics.md?metric=VBench%20Semantic%20Score>): Higher is better; unit: PERCENT.

- [VBench Perceptual Quality Score](<https://www.sota2.com/research/sota/text-to-video-generation-on-vbench-quality-and-efficiency-metrics.md?metric=VBench%20Perceptual%20Quality%20Score>): Higher is better; unit: PERCENT.

- [VBench Total Score](<https://www.sota2.com/research/sota/text-to-video-generation-on-vbench-quality-and-efficiency-metrics.md?metric=VBench%20Total%20Score>): Higher is better; unit: PERCENT.

- [Sparsity](<https://www.sota2.com/research/sota/text-to-video-generation-on-vbench-quality-and-efficiency-metrics.md?metric=Sparsity>): Higher is better; unit: PERCENT.

- [Latency (s)](<https://www.sota2.com/research/sota/text-to-video-generation-on-vbench-quality-and-efficiency-metrics.md?metric=Latency%20(s)>): Lower is better; unit: TIME.

- [Speedup Factor](<https://www.sota2.com/research/sota/text-to-video-generation-on-vbench-quality-and-efficiency-metrics.md?metric=Speedup%20Factor>): Higher is better; unit: SCALAR.

- [VBench Score](<https://www.sota2.com/research/sota/text-to-video-generation-on-vbench-quality-and-efficiency-metrics.md?metric=VBench%20Score>): Higher is better; unit: PERCENT.

## Recorded results

| Method | VBench Semantic Score | Result date | Report paper |
| --- | --- | --- | --- |
| CalibAtt | 77.17 | Not recorded | 69aa5f9c2d603e6de3e41a9a |
| RadialAttention | 77.02 | Not recorded | 69aa5f9c2d603e6de3e41a9a |
| SpargeAttention | 76.4 | Not recorded | 69aa5f9c2d603e6de3e41a9a |
| SVG2 | 76.19 | Not recorded | 69aa5f9c2d603e6de3e41a9a |
| Dense (FA3) | 76.06 | Not recorded | 69aa5f9c2d603e6de3e41a9a |
| CalibAtt | 74.04 | Not recorded | 69aa5f9c2d603e6de3e41a9a |
| SpargeAttention | 73.47 | Not recorded | 69aa5f9c2d603e6de3e41a9a |
| RadialAttention | 73.24 | Not recorded | 69aa5f9c2d603e6de3e41a9a |
| SVG2 | 73.19 | Not recorded | 69aa5f9c2d603e6de3e41a9a |
| Dense (FA3) | 73.17 | Not recorded | 69aa5f9c2d603e6de3e41a9a |

### CalibAtt

VBench Semantic Score: 77.17. Result ID: 69aaaa086dbb7075b954f119.

[Reporting paper](<https://www.sota2.com/research/paper/69aa5f9c2d603e6de3e41a9a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69aa5f9c2d603e6de3e41a9a.md>)

Recorded evaluation configuration: {"Resolution":"480p","Num Frames":"81","Sampling Steps":"4","Backbone":"LightX2V"}. This record does not assert independent reproduction.

### RadialAttention

VBench Semantic Score: 77.02. Result ID: 69aaaa086dbb7075b954f117.

[Reporting paper](<https://www.sota2.com/research/paper/69aa5f9c2d603e6de3e41a9a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79ab1e209131cea7ecf7.md>)

Recorded evaluation configuration: {"Resolution":"480p","Num Frames":"81","Sampling Steps":"4","Backbone":"LightX2V"}. This record does not assert independent reproduction.

### SpargeAttention

VBench Semantic Score: 76.4. Result ID: 69aaaa086dbb7075b954f116.

[Reporting paper](<https://www.sota2.com/research/paper/69aa5f9c2d603e6de3e41a9a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7bbe1e209131cea8a28a.md>)

Recorded evaluation configuration: {"Resolution":"480p","Num Frames":"81","Sampling Steps":"4","Backbone":"LightX2V"}. This record does not assert independent reproduction.

### SVG2

VBench Semantic Score: 76.19. Result ID: 69aaaa086dbb7075b954f118.

[Reporting paper](<https://www.sota2.com/research/paper/69aa5f9c2d603e6de3e41a9a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7be31e209131cea8b13e.md>)

Recorded evaluation configuration: {"Resolution":"480p","Num Frames":"81","Sampling Steps":"4","Backbone":"LightX2V"}. This record does not assert independent reproduction.

### Dense (FA3)

VBench Semantic Score: 76.06. Result ID: 69aaaa086dbb7075b954f115.

[Reporting paper](<https://www.sota2.com/research/paper/69aa5f9c2d603e6de3e41a9a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa740f1e209131cea5f89c.md>)

[Implementation](<https://github.com/Dao-AILab/flash-attention>)

Recorded evaluation configuration: {"Resolution":"480p","Num Frames":"81","Sampling Steps":"4","Backbone":"LightX2V"}. This record does not assert independent reproduction.

### CalibAtt

VBench Semantic Score: 74.04. Result ID: 69aaaa086dbb7075b954f11e.

[Reporting paper](<https://www.sota2.com/research/paper/69aa5f9c2d603e6de3e41a9a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69aa5f9c2d603e6de3e41a9a.md>)

Recorded evaluation configuration: {"Resolution":"720p","Num Frames":"81","Sampling Steps":"4","Backbone":"LightX2V"}. This record does not assert independent reproduction.

### SpargeAttention

VBench Semantic Score: 73.47. Result ID: 69aaaa086dbb7075b954f11b.

[Reporting paper](<https://www.sota2.com/research/paper/69aa5f9c2d603e6de3e41a9a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7bbe1e209131cea8a28a.md>)

Recorded evaluation configuration: {"Resolution":"720p","Num Frames":"81","Sampling Steps":"4","Backbone":"LightX2V"}. This record does not assert independent reproduction.

### RadialAttention

VBench Semantic Score: 73.24. Result ID: 69aaaa086dbb7075b954f11c.

[Reporting paper](<https://www.sota2.com/research/paper/69aa5f9c2d603e6de3e41a9a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79ab1e209131cea7ecf7.md>)

Recorded evaluation configuration: {"Resolution":"720p","Num Frames":"81","Sampling Steps":"4","Backbone":"LightX2V"}. This record does not assert independent reproduction.

### SVG2

VBench Semantic Score: 73.19. Result ID: 69aaaa086dbb7075b954f11d.

[Reporting paper](<https://www.sota2.com/research/paper/69aa5f9c2d603e6de3e41a9a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7be31e209131cea8b13e.md>)

Recorded evaluation configuration: {"Resolution":"720p","Num Frames":"81","Sampling Steps":"4","Backbone":"LightX2V"}. This record does not assert independent reproduction.

### Dense (FA3)

VBench Semantic Score: 73.17. Result ID: 69aaaa086dbb7075b954f11a.

[Reporting paper](<https://www.sota2.com/research/paper/69aa5f9c2d603e6de3e41a9a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa740f1e209131cea5f89c.md>)

[Implementation](<https://github.com/Dao-AILab/flash-attention>)

Recorded evaluation configuration: {"Resolution":"720p","Num Frames":"81","Sampling Steps":"4","Backbone":"LightX2V"}. This record does not assert independent reproduction.

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