Salvato in:
| Autori principali: | , , , , , , , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2026
|
| Soggetti: | |
| Accesso online: | https://arxiv.org/abs/2605.23176 |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866918517946712064 |
|---|---|
| author | Vo, Hao Vo, Khoa Nguyen, Phu Loc Tran, Sieu Nguyen, Duc Minh Cuong, Ngo Xuan Gawugah, Gladys Godavarthi, Sreevenkata Anjani Tishita Rainwater, Chase Bui, Nghi D. Q. Nguyen, Anh Nguyen, Duy Minh Ho Le, Ngan |
| author_facet | Vo, Hao Vo, Khoa Nguyen, Phu Loc Tran, Sieu Nguyen, Duc Minh Cuong, Ngo Xuan Gawugah, Gladys Godavarthi, Sreevenkata Anjani Tishita Rainwater, Chase Bui, Nghi D. Q. Nguyen, Anh Nguyen, Duy Minh Ho Le, Ngan |
| contents | Spatiotemporal intelligence in autonomous driving (AD) requires an agent to integrate multi-view observations into a coherent scene representation, maintain object continuity across viewpoints and time, and reason about spatial relations, interactions, and future dynamics. However, existing AD vision-language benchmarks largely focus on single-view, static, ego-centric, or single-source question answering, leaving it unclear whether current Vision-Language Models (VLMs) can truly construct and reason over dynamic driving scenes. We introduce DriveSpatial, a benchmark of 15.6K human-verified QA pairs across 20 tasks from five large-scale AD datasets. DriveSpatial evaluates four abilities: Cognitive Scene Construction, Multi-view Relational Understanding, Temporal Reasoning, and Generalization. Unlike prior benchmarks, DriveSpatial is generated from a dynamic multi-relational scene graph that encodes object states, spatial relations, interactions, camera visibility, and temporal correspondences, enabling QA pairs that enforce genuine cross-view and spatiotemporal reasoning. Evaluating 15 representative VLMs reveals a substantial human-model gap: the strongest model trails humans by 28.4 points, with Cognitive Scene Construction emerging as the key bottleneck. Further diagnostics show that language-only prompting is insufficient, while explicit BEV grounding consistently improves performance. These results suggest that current VLMs lack the scene-construction ability needed for reliable spatiotemporal driving intelligence. DriveSpatial and its construction pipeline will be released to support future research. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_23176 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | DRIVESPATIAL: A Benchmark for Spatiotemporal Intelligence in VLMs for Autonomous Driving Vo, Hao Vo, Khoa Nguyen, Phu Loc Tran, Sieu Nguyen, Duc Minh Cuong, Ngo Xuan Gawugah, Gladys Godavarthi, Sreevenkata Anjani Tishita Rainwater, Chase Bui, Nghi D. Q. Nguyen, Anh Nguyen, Duy Minh Ho Le, Ngan Computer Vision and Pattern Recognition Spatiotemporal intelligence in autonomous driving (AD) requires an agent to integrate multi-view observations into a coherent scene representation, maintain object continuity across viewpoints and time, and reason about spatial relations, interactions, and future dynamics. However, existing AD vision-language benchmarks largely focus on single-view, static, ego-centric, or single-source question answering, leaving it unclear whether current Vision-Language Models (VLMs) can truly construct and reason over dynamic driving scenes. We introduce DriveSpatial, a benchmark of 15.6K human-verified QA pairs across 20 tasks from five large-scale AD datasets. DriveSpatial evaluates four abilities: Cognitive Scene Construction, Multi-view Relational Understanding, Temporal Reasoning, and Generalization. Unlike prior benchmarks, DriveSpatial is generated from a dynamic multi-relational scene graph that encodes object states, spatial relations, interactions, camera visibility, and temporal correspondences, enabling QA pairs that enforce genuine cross-view and spatiotemporal reasoning. Evaluating 15 representative VLMs reveals a substantial human-model gap: the strongest model trails humans by 28.4 points, with Cognitive Scene Construction emerging as the key bottleneck. Further diagnostics show that language-only prompting is insufficient, while explicit BEV grounding consistently improves performance. These results suggest that current VLMs lack the scene-construction ability needed for reliable spatiotemporal driving intelligence. DriveSpatial and its construction pipeline will be released to support future research. |
| title | DRIVESPATIAL: A Benchmark for Spatiotemporal Intelligence in VLMs for Autonomous Driving |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2605.23176 |