Research map / Vision-language and multimodal / Multimodal LLMs
Kilometer-Vision: A New Frontier for Large-Scale Spatial Awareness in VLMs
Aravindh Mahendran, Michael King, Matthew Grimes, Antoine Yang, Tyler Zhu, Joseph Heyward, Tengda Han, Shiry Ginosar, Chen Sun, Dima Damen, Simon Osindero, Noah Snavely and 3 more
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 |
|---|---|---|---|
| increase frames from 32 to 600 (ablation) | Gemini 2.5 Flash · KilometerVision | accuracy: 32 -> 600; ~38% accuracy | No clear effect |
| increase frames from 32 to 500 (ablation) | GPT-5 · KilometerVision | accuracy: 45% to 57% | Improved |
| halve frame resolution from 400 to 200 pixels (ablation) | Claude Opus · KilometerVision | accuracy | Hurt |
| add map with text labels (ablation) | Gemini 2.5 Flash · KilometerVision route summary (Gemini 2.5 Flash) | top-1 accuracy: 42.2 -> 49.0 | Improved |
| add map without text labels (ablation) | Gemini 2.5 Flash · KilometerVision route summary (Gemini 2.5 Flash) | top-1 accuracy: 42.2 -> 37.4 | Hurt |
| remove the video and keep labeled map (ablation) | Gemini 2.5 Flash · KilometerVision route summary (Gemini 2.5 Flash) | top-1 accuracy: 49.0 -> 53.1 | Improved |
| add ground-truth route summary text (ablation) | Gemini 2.5 Flash · KilometerVision map trace with text (Gemini 2.5 Flash) | top-1 accuracy: 28.6 -> 34.7 | Improved |
| add ground-truth route summary text (ablation) | Gemini 2.5 Flash · KilometerVision map trace without text (Gemini 2.5 Flash) | top-1 accuracy: 28.6 -> 26.5 | Hurt |
| remove the video and keep route summary (ablation) | Gemini 2.5 Flash · KilometerVision map trace with text (Gemini 2.5 Flash) | top-1 accuracy: 34.7 -> 46.3 | Improved |
| remove the video and keep route summary (ablation) | Gemini 2.5 Flash · KilometerVision map trace without text (Gemini 2.5 Flash) | top-1 accuracy: 26.5 -> 38.1 | Improved |
| add camera pose or VPS coordinates (ablation) | Gemini 2.5 Flash · KilometerVision map trace | accuracy | No clear effect |
| enable thinking (ablation) | Claude Opus · KilometerVision | accuracy: 34.9% vs 35.1% | No clear effect |
Closest papers by meaning
- Is A Picture Worth A Thousand Words? Delving Into Spatial Reasoning for Vision Language Models (NeurIPS 2024 · similarity 0.91)
- Spatial-DISE: A Unified Benchmark for Evaluating Spatial Reasoning in Vision-Language Models (ICLR 2026 · similarity 0.91)
- SD-VLM: Spatial Measuring and Understanding with Depth-Encoded Vision-Language Models (NeurIPS 2025 · similarity 0.91)
- From Indoor to Open World: Revealing the Spatial Reasoning Gap in MLLMs (CVPR 2026 · similarity 0.91)
- Spatial-MLLM: Boosting MLLM Capabilities in Visual-based Spatial Intelligence (NeurIPS 2025 · similarity 0.90)
- SpatialScore: Towards Comprehensive Evaluation for Spatial Intelligence (CVPR 2026 · similarity 0.90)
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.
