# Spatio-temporal reasoning on VSTemporalI-Bench

Source: [Spatio-temporal reasoning on VSTemporalI-Bench](<https://www.sota2.com/research/sota/spatio-temporal-reasoning-on-vstemporali-bench>)

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

Data updated: 2026-03-31T09:10:40.612Z

Evaluation of model performance on the VSTemporalI-Bench dataset, focusing on Spatio-temporal reasoning capabilities, measured by average error and specific camera/object distance 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 | Spatio-temporal reasoning |
| Dataset (including any stated split) | VSTemporalI-Bench |
| Metric | Object-Object Relative Position Error |
| Unit | SCALAR |
| Direction | Lower is better |
| Results loaded | 14 of 14 |
| 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

- [Object-Object Relative Position Error](<https://www.sota2.com/research/sota/spatio-temporal-reasoning-on-vstemporali-bench.md?metric=Object-Object%20Relative%20Position%20Error>): Lower is better; unit: SCALAR.

- [Average Error](<https://www.sota2.com/research/sota/spatio-temporal-reasoning-on-vstemporali-bench.md?metric=Average%20Error>): Lower is better; unit: SCALAR.

- [Camera-Object Absolute Distance](<https://www.sota2.com/research/sota/spatio-temporal-reasoning-on-vstemporali-bench.md?metric=Camera-Object%20Absolute%20Distance>): Lower is better; unit: SCALAR.

- [Camera Displacement](<https://www.sota2.com/research/sota/spatio-temporal-reasoning-on-vstemporali-bench.md?metric=Camera%20Displacement>): Lower is better; unit: SCALAR.

- [Camera Movement Direction Error](<https://www.sota2.com/research/sota/spatio-temporal-reasoning-on-vstemporali-bench.md?metric=Camera%20Movement%20Direction%20Error>): Lower is better; unit: SCALAR.

- [Camera-Object Relative Distance](<https://www.sota2.com/research/sota/spatio-temporal-reasoning-on-vstemporali-bench.md?metric=Camera-Object%20Relative%20Distance>): Lower is better; unit: SCALAR.

## Recorded results

| Method | Object-Object Relative Position Error | Result date | Report paper |
| --- | --- | --- | --- |
| Chance Level (Random) | 36.1 | Not recorded | 68fa790d1e209131cea7be1f |
| VILA-1.5-8B | 50.4 | Not recorded | 68fa790d1e209131cea7be1f |
| Chance Level (Frequency) | 52.2 | Not recorded | 68fa790d1e209131cea7be1f |
| LongVILA-8B | 52.3 | Not recorded | 68fa790d1e209131cea7be1f |
| LongVA-7B | 57.9 | Not recorded | 68fa790d1e209131cea7be1f |
| GPT-4o | 58.1 | Not recorded | 68fa790d1e209131cea7be1f |
| LLaVA-OneVision-7B | 62.1 | Not recorded | 68fa790d1e209131cea7be1f |
| LLaVA-NeXT-Video-7B | 64.7 | Not recorded | 68fa790d1e209131cea7be1f |
| VILA-1.5-40B | 65.4 | Not recorded | 68fa790d1e209131cea7be1f |
| InternVL2-8B | 68 | Not recorded | 68fa790d1e209131cea7be1f |
| LLaVA-NeXT-Video-72B | 78.3 | Not recorded | 68fa790d1e209131cea7be1f |
| Gemini-2.5 Pro | 84 | Not recorded | 68fa790d1e209131cea7be1f |
| VLM-3R (7B) | 86.5 | Not recorded | 68fa790d1e209131cea7be1f |
| †Human Level | 97.5 | Not recorded | 68fa790d1e209131cea7be1f |

### Chance Level (Random)

Object-Object Relative Position Error: 36.1. Result ID: 69cb8ddc47f34ce924b409fc.

[Reporting paper](<https://www.sota2.com/research/paper/68fa790d1e209131cea7be1f.md>)

Recorded evaluation configuration: {"Rank":"-"}. This record does not assert independent reproduction.

### VILA-1.5-8B

Object-Object Relative Position Error: 50.4. Result ID: 69cb8ddc47f34ce924b40a06.

[Reporting paper](<https://www.sota2.com/research/paper/68fa790d1e209131cea7be1f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73311e209131cea5b06b.md>)

[Implementation](<https://github.com/NVlabs/VILA>)

Recorded evaluation configuration: {"Rank":"8","Model Type":"Open-sourced VLMs"}. This record does not assert independent reproduction.

### Chance Level (Frequency)

Object-Object Relative Position Error: 52.2. Result ID: 69cb8ddc47f34ce924b409fd.

[Reporting paper](<https://www.sota2.com/research/paper/68fa790d1e209131cea7be1f.md>)

Recorded evaluation configuration: {"Rank":"-"}. This record does not assert independent reproduction.

### LongVILA-8B

Object-Object Relative Position Error: 52.3. Result ID: 69cb8ddc47f34ce924b40a05.

[Reporting paper](<https://www.sota2.com/research/paper/68fa790d1e209131cea7be1f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa76271e209131cea6b860.md>)

[Implementation](<https://www.sota2.com/NVlabs/LongVILA>)

Recorded evaluation configuration: {"Rank":"11","Model Type":"Open-sourced VLMs"}. This record does not assert independent reproduction.

### LongVA-7B

Object-Object Relative Position Error: 57.9. Result ID: 69cb8ddc47f34ce924b40a03.

[Reporting paper](<https://www.sota2.com/research/paper/68fa790d1e209131cea7be1f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73ea1e209131cea5edfe.md>)

[Implementation](<https://github.com/EvolvingLMMs-Lab/LongVA>)

Recorded evaluation configuration: {"Rank":"10","Model Type":"Open-sourced VLMs"}. This record does not assert independent reproduction.

### GPT-4o

Object-Object Relative Position Error: 58.1. Result ID: 69cb8ddc47f34ce924b40a00.

[Reporting paper](<https://www.sota2.com/research/paper/68fa790d1e209131cea7be1f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75681e209131cea67a98.md>)

Recorded evaluation configuration: {"Rank":"2","Model Type":"Proprietary Models (API)"}. This record does not assert independent reproduction.

### LLaVA-OneVision-7B

Object-Object Relative Position Error: 62.1. Result ID: 69cb8ddc47f34ce924b40a02.

[Reporting paper](<https://www.sota2.com/research/paper/68fa790d1e209131cea7be1f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa756f1e209131cea67e58.md>)

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

Recorded evaluation configuration: {"Rank":"4","Model Type":"Open-sourced VLMs"}. This record does not assert independent reproduction.

### LLaVA-NeXT-Video-7B

Object-Object Relative Position Error: 64.7. Result ID: 69cb8ddc47f34ce924b40a01.

[Reporting paper](<https://www.sota2.com/research/paper/68fa790d1e209131cea7be1f.md>)

Recorded evaluation configuration: {"Rank":"5","Model Type":"Open-sourced VLMs"}. This record does not assert independent reproduction.

### VILA-1.5-40B

Object-Object Relative Position Error: 65.4. Result ID: 69cb8ddc47f34ce924b40a07.

[Reporting paper](<https://www.sota2.com/research/paper/68fa790d1e209131cea7be1f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73311e209131cea5b06b.md>)

[Implementation](<https://github.com/NVlabs/VILA>)

Recorded evaluation configuration: {"Rank":"6","Model Type":"Open-sourced VLMs"}. This record does not assert independent reproduction.

### InternVL2-8B

Object-Object Relative Position Error: 68. Result ID: 69cb8ddc47f34ce924b40a04.

[Reporting paper](<https://www.sota2.com/research/paper/68fa790d1e209131cea7be1f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73081e209131cea59f8b.md>)

[Implementation](<https://github.com/OpenGVLab/InternVL>)

Recorded evaluation configuration: {"Rank":"3","Model Type":"Open-sourced VLMs"}. This record does not assert independent reproduction.

### LLaVA-NeXT-Video-72B

Object-Object Relative Position Error: 78.3. Result ID: 69cb8ddc47f34ce924b40a08.

[Reporting paper](<https://www.sota2.com/research/paper/68fa790d1e209131cea7be1f.md>)

Recorded evaluation configuration: {"Rank":"2","Model Type":"Open-sourced VLMs"}. This record does not assert independent reproduction.

### Gemini-2.5 Pro

Object-Object Relative Position Error: 84. Result ID: 69cb8ddc47f34ce924b409ff.

[Reporting paper](<https://www.sota2.com/research/paper/68fa790d1e209131cea7be1f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa76361e209131cea6be11.md>)

Recorded evaluation configuration: {"Rank":"1","Model Type":"Proprietary Models (API)"}. This record does not assert independent reproduction.

### VLM-3R (7B)

Object-Object Relative Position Error: 86.5. Result ID: 69cb8ddc47f34ce924b40a09.

[Reporting paper](<https://www.sota2.com/research/paper/68fa790d1e209131cea7be1f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa790d1e209131cea7be1f.md>)

Recorded evaluation configuration: {"Rank":"1","Model Type":"Open-sourced VLMs"}. This record does not assert independent reproduction.

### †Human Level

Object-Object Relative Position Error: 97.5. Result ID: 69cb8ddc47f34ce924b409fe.

[Reporting paper](<https://www.sota2.com/research/paper/68fa790d1e209131cea7be1f.md>)

Recorded evaluation configuration: {"Rank":"-"}. This record does not assert independent reproduction.

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