Research map / Vision-language and multimodal / Multimodal LLMs
CrossView: Can Vision-Language Models Reason Across Cameras?
Sahil Shah, S P Sharan, Harsh Goel, Manvik Pasula, Adithya Hebbalae, Minkyu Choi, Sandeep Chinchali
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| scale up model or use proprietary GPT-5.2 on multi-camera questions | VLMs on CrossView · CrossView | accuracy: GPT-5.2 25.4 on nuScenes temporal; 34.6 on Ego-Exo4D best camera | No clear effect |
| use all six cameras instead of the front camera only (ablation) | Qwen2.5-7B · CrossView nuScenes counting | accuracy (%): 18.7 -> 27.5 | Improved |
| use all six cameras instead of the front camera only (ablation) | Qwen2.5-7B · CrossView nuScenes spatial | accuracy (%): 41.5 -> 45.9 | Improved |
| use all six cameras instead of the front camera only (ablation) | Qwen2.5-7B · CrossView nuScenes event ordering | accuracy (%): 32.0 (front) vs 24.4 (all cameras) | Hurt |
| use all cameras instead of ego or best exo camera (ablation) | Qwen2.5-7B · CrossView Ego-Exo4D temporal | accuracy (%): 40.4-40.7 -> 45.2 | Improved |
| stitched tiling instead of uniform per-camera sampling (ablation) | Qwen2.5-7B · CrossView nuScenes | accuracy (%): counting 27.5 -> 38.0; event ordering 24.4 -> 39.6 | Improved |
| stitched tiling instead of uniform per-camera sampling (ablation) | Qwen2.5-7B · CrossView MEVA | accuracy (%): spatial 17.6 -> 34.2; event ordering 83.2 -> 70.9 | Mixed |
| stitched tiling instead of uniform per-camera sampling (ablation) | Qwen2.5-7B · CrossView AgiBot | accuracy (%): event ordering 54.4 -> 48.4 | Hurt |
| stitched tiling instead of uniform sampling (ablation) | Qwen2.5-7B · CrossView Ego-Exo4D | accuracy (%): temporal 45.2 -> 53.6; event ordering 49.2 -> 48.0 | Mixed |
Closest papers by meaning
- MVVBench: Benchmarking 4D Reasoning in Vision-Language Models (NeurIPS 2026 · similarity 0.90)
- PerceptionLM: Open-Access Data and Models for Detailed Visual Understanding (NeurIPS 2025 · similarity 0.88)
- OmniVideoBench: Towards Audio-Visual Understanding Evaluation for Omni MLLMs (ICLR 2026 · similarity 0.87)
- Black Swan: Abductive and Defeasible Video Reasoning in Unpredictable Events (CVPR 2025 · similarity 0.87)
- ROVER: Recursive Reasoning Over Videos with Vision-Language Models for Embodied Tasks (NeurIPS 2025 · similarity 0.87)
- Seeing from Another Perspective: Evaluating Multi-View Understanding in MLLMs (AAAI 2026 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
