# Video Generation on Temporal Order Hard Set, 3-4 Steps

Source: [Video Generation on Temporal Order Hard Set, 3-4 Steps](<https://www.sota2.com/research/sota/video-generation-on-temporal-order-hard-set-3-4-steps>)

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

Data updated: 2026-06-29T07:18:17.841Z

Evaluation of video generation models specifically on the Temporal Order Hard Set (3-4 Steps subset), measured using AI judge scores for temporal following accuracy.

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 | Video Generation |
| Dataset (including any stated split) | Temporal Order Hard Set, 3-4 Steps |
| Metric | Temporal Following Score (Qwen Judge) |
| Unit | SCALAR |
| Direction | Higher is better |
| Results loaded | 7 of 7 |
| 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

- [Temporal Following Score (Qwen Judge)](<https://www.sota2.com/research/sota/video-generation-on-temporal-order-hard-set-3-4-steps.md?metric=Temporal%20Following%20Score%20(Qwen%20Judge)>): Higher is better; unit: SCALAR.

- [Temporal Following Score (Gemini Judge)](<https://www.sota2.com/research/sota/video-generation-on-temporal-order-hard-set-3-4-steps.md?metric=Temporal%20Following%20Score%20(Gemini%20Judge)>): Higher is better; unit: SCALAR.

## Recorded results

| Method | Temporal Following Score (Qwen Judge) | Result date | Report paper |
| --- | --- | --- | --- |
| TempAct | 0.352 | Not recorded | 6a41fbf3206542251d7cca47 |
| TempAct | 0.336 | Not recorded | 6a41fbf3206542251d7cca47 |
| LongLive | 0.314 | Not recorded | 6a41fbf3206542251d7cca47 |
| Casual Forcing | 0.304 | Not recorded | 6a41fbf3206542251d7cca47 |
| LongLive | 0.302 | Not recorded | 6a41fbf3206542251d7cca47 |
| Self-Forcing | 0.269 | Not recorded | 6a41fbf3206542251d7cca47 |
| Self-Forcing | 0.24 | Not recorded | 6a41fbf3206542251d7cca47 |

### TempAct

Temporal Following Score (Qwen Judge): 0.352. Result ID: 6a421c31fbc373592c872d2c.

[Reporting paper](<https://www.sota2.com/research/paper/6a41fbf3206542251d7cca47.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a41fbf3206542251d7cca47.md>)

Recorded evaluation configuration: {"Prompting Strategy":"Step-prompt video generation","Backbone":"LongLive"}. This record does not assert independent reproduction.

### TempAct

Temporal Following Score (Qwen Judge): 0.336. Result ID: 6a421c31fbc373592c872d2a.

[Reporting paper](<https://www.sota2.com/research/paper/6a41fbf3206542251d7cca47.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a41fbf3206542251d7cca47.md>)

Recorded evaluation configuration: {"Prompting Strategy":"Step-prompt video generation","Backbone":"Self-Forcing"}. This record does not assert independent reproduction.

### LongLive

Temporal Following Score (Qwen Judge): 0.314. Result ID: 6a421c31fbc373592c872d2b.

[Reporting paper](<https://www.sota2.com/research/paper/6a41fbf3206542251d7cca47.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7c251e209131cea8cb13.md>)

Recorded evaluation configuration: {"Prompting Strategy":"Step-prompt video generation"}. This record does not assert independent reproduction.

### Casual Forcing

Temporal Following Score (Qwen Judge): 0.304. Result ID: 6a421c31fbc373592c872d27.

[Reporting paper](<https://www.sota2.com/research/paper/6a41fbf3206542251d7cca47.md>)

[Method origin paper](<https://www.sota2.com/research/paper/698181ada8b01e76cc8e27d4.md>)

[Implementation](<https://github.com/thu-ml/Causal-Forcing>)

Recorded evaluation configuration: {"Prompting Strategy":"Single-prompt video generation"}. This record does not assert independent reproduction.

### LongLive

Temporal Following Score (Qwen Judge): 0.302. Result ID: 6a421c31fbc373592c872d28.

[Reporting paper](<https://www.sota2.com/research/paper/6a41fbf3206542251d7cca47.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7c251e209131cea8cb13.md>)

Recorded evaluation configuration: {"Prompting Strategy":"Single-prompt video generation"}. This record does not assert independent reproduction.

### Self-Forcing

Temporal Following Score (Qwen Judge): 0.269. Result ID: 6a421c31fbc373592c872d29.

[Reporting paper](<https://www.sota2.com/research/paper/6a41fbf3206542251d7cca47.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa793a1e209131cea7cb5a.md>)

Recorded evaluation configuration: {"Prompting Strategy":"Step-prompt video generation"}. This record does not assert independent reproduction.

### Self-Forcing

Temporal Following Score (Qwen Judge): 0.24. Result ID: 6a421c31fbc373592c872d26.

[Reporting paper](<https://www.sota2.com/research/paper/6a41fbf3206542251d7cca47.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa793a1e209131cea7cb5a.md>)

Recorded evaluation configuration: {"Prompting Strategy":"Single-prompt video generation"}. This record does not assert independent reproduction.

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