# HOI Video Generation on in-the-wild dataset (test)

Source: [HOI Video Generation on in-the-wild dataset (test)](<https://www.sota2.com/research/sota/hoi-video-generation-on-in-the-wild-dataset-test>)

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

Data updated: 2026-02-26T05:45:34.952Z

Evaluation of Human-Object Interaction (HOI) video generation models on the test split of the in-the-wild dataset, assessed using metrics covering fidelity, temporal consistency, alignment, and overall quality.

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 | HOI Video Generation |
| Dataset (including any stated split) | in-the-wild dataset (test) |
| Metric | Fréchet Video Distance (FVD) |
| Unit | SCALAR |
| Direction | Lower is better |
| Results loaded | 4 of 4 |
| 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

- [Fréchet Video Distance (FVD)](<https://www.sota2.com/research/sota/hoi-video-generation-on-in-the-wild-dataset-test.md?metric=Fr%C3%A9chet%20Video%20Distance%20(FVD)>): Lower is better; unit: SCALAR.

- [Temporal Error (TE)](<https://www.sota2.com/research/sota/hoi-video-generation-on-in-the-wild-dataset-test.md?metric=Temporal%20Error%20(TE)>): Lower is better; unit: SCALAR.

- [Alignment Score (CA)](<https://www.sota2.com/research/sota/hoi-video-generation-on-in-the-wild-dataset-test.md?metric=Alignment%20Score%20(CA)>): Higher is better; unit: PERCENT.

- [Clip Similarity (CLIPSIM)](<https://www.sota2.com/research/sota/hoi-video-generation-on-in-the-wild-dataset-test.md?metric=Clip%20Similarity%20(CLIPSIM)>): Higher is better; unit: PERCENT.

- [Spatial Consistency (SC)](<https://www.sota2.com/research/sota/hoi-video-generation-on-in-the-wild-dataset-test.md?metric=Spatial%20Consistency%20(SC)>): Higher is better; unit: PERCENT.

- [Bounding Box Consistency (BC)](<https://www.sota2.com/research/sota/hoi-video-generation-on-in-the-wild-dataset-test.md?metric=Bounding%20Box%20Consistency%20(BC)>): Higher is better; unit: PERCENT.

- [Deviation/Diversity Score (DD)](<https://www.sota2.com/research/sota/hoi-video-generation-on-in-the-wild-dataset-test.md?metric=Deviation%2FDiversity%20Score%20(DD)>): Higher is better; unit: PERCENT.

- [Motion Score (MS)](<https://www.sota2.com/research/sota/hoi-video-generation-on-in-the-wild-dataset-test.md?metric=Motion%20Score%20(MS)>): Higher is better; unit: PERCENT.

- [Appearance Quality (AQ)](<https://www.sota2.com/research/sota/hoi-video-generation-on-in-the-wild-dataset-test.md?metric=Appearance%20Quality%20(AQ)>): Higher is better; unit: PERCENT.

- [Interaction Quality (IQ)](<https://www.sota2.com/research/sota/hoi-video-generation-on-in-the-wild-dataset-test.md?metric=Interaction%20Quality%20(IQ)>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Fréchet Video Distance (FVD) | Result date | Report paper |
| --- | --- | --- | --- |
| VHOI | 484 | Not recorded | 693a4ffc6cfd488761a549c0 |
| TORA | 606 | Not recorded | 693a4ffc6cfd488761a549c0 |
| TORA | 726 | Not recorded | 693a4ffc6cfd488761a549c0 |
| Go-with-the-flow | 780 | Not recorded | 693a4ffc6cfd488761a549c0 |

### VHOI

Fréchet Video Distance (FVD): 484. Result ID: 699863705698665ec4c26c98.

[Reporting paper](<https://www.sota2.com/research/paper/693a4ffc6cfd488761a549c0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/693a4ffc6cfd488761a549c0.md>)

Recorded evaluation configuration: {"Mode":"Default"}. This record does not assert independent reproduction.

### TORA

Fréchet Video Distance (FVD): 606. Result ID: 699863705698665ec4c26c96.

[Reporting paper](<https://www.sota2.com/research/paper/693a4ffc6cfd488761a549c0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa777c1e209131cea737ec.md>)

Recorded evaluation configuration: {"Mode":"Finetuned"}. This record does not assert independent reproduction.

### TORA

Fréchet Video Distance (FVD): 726. Result ID: 699863705698665ec4c26c95.

[Reporting paper](<https://www.sota2.com/research/paper/693a4ffc6cfd488761a549c0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa777c1e209131cea737ec.md>)

Recorded evaluation configuration: {"Mode":"Default"}. This record does not assert independent reproduction.

### Go-with-the-flow

Fréchet Video Distance (FVD): 780. Result ID: 699863705698665ec4c26c97.

[Reporting paper](<https://www.sota2.com/research/paper/693a4ffc6cfd488761a549c0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7a651e209131cea82e5f.md>)

[Implementation](<https://eyeline-labs.github.io/Go-with-the-Flow/>)

Recorded evaluation configuration: {"Mode":"Default"}. This record does not assert independent reproduction.

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