# Direct long-horizon prediction on SACSoN (4-second future horizon)

Source: [Direct long-horizon prediction on SACSoN (4-second future horizon)](<https://www.sota2.com/research/sota/direct-long-horizon-prediction-on-sacson-4-second-future-horizon>)

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

Data updated: 2026-03-11T08:22:59.238Z

Evaluation of direct long-horizon prediction performance on the SACSoN dataset, specifically focusing on the 4-second future horizon setting, using perceptual and generative quality metrics.

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 | Direct long-horizon prediction |
| Dataset (including any stated split) | SACSoN (4-second future horizon) |
| Metric | LPIPS |
| Unit | SCALAR |
| Direction | Lower is better |
| Results loaded | 2 of 2 |
| 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

- [LPIPS](<https://www.sota2.com/research/sota/direct-long-horizon-prediction-on-sacson-4-second-future-horizon.md?metric=LPIPS>): Lower is better; unit: SCALAR.

- [DreamSim](<https://www.sota2.com/research/sota/direct-long-horizon-prediction-on-sacson-4-second-future-horizon.md?metric=DreamSim>): Higher is better; unit: SCALAR.

- [DINO Distance](<https://www.sota2.com/research/sota/direct-long-horizon-prediction-on-sacson-4-second-future-horizon.md?metric=DINO%20Distance>): Lower is better; unit: SCALAR.

- [FID](<https://www.sota2.com/research/sota/direct-long-horizon-prediction-on-sacson-4-second-future-horizon.md?metric=FID>): Lower is better; unit: SCALAR.

## Recorded results

| Method | LPIPS | Result date | Report paper |
| --- | --- | --- | --- |
| RAE-NWM | 0.303 | Not recorded | 69b0f7232d603e6de3e45e7e |
| NWM | 0.407 | Not recorded | 69b0f7232d603e6de3e45e7e |

### RAE-NWM

LPIPS: 0.303. Result ID: 69b126486dbb7075b95b7afd.

[Reporting paper](<https://www.sota2.com/research/paper/69b0f7232d603e6de3e45e7e.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69b0f7232d603e6de3e45e7e.md>)

Recorded evaluation configuration: {"Prediction Horizon":"4s"}. This record does not assert independent reproduction.

### NWM

LPIPS: 0.407. Result ID: 69b126486dbb7075b95b7afc.

[Reporting paper](<https://www.sota2.com/research/paper/69b0f7232d603e6de3e45e7e.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa77ea1e209131cea75f91.md>)

[Implementation](<https://amirbar.net/nwm>)

Recorded evaluation configuration: {"Prediction Horizon":"4s"}. This record does not assert independent reproduction.

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