AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

UrbanLLaVA: A Multi-modal Large Language Model for Urban Intelligence

Jie Feng, Shengyuan Wang, Tianhui Liu, Yanxin Xi, Yong Li

ICCV 2025 · Vision-language and multimodal › Multimodal LLMs · approach: Benchmarks and analysis of MLLMs · 29 citations (OpenAlex)

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 changedWhereMeasuredEffect
fine-tune VILA1.5-8B with UData and UTrain on urban dataVILA1.5-8B · UBench Beijing Geo+Trajaccuracy: 0.0589 -> 0.2800Improved
UrbanLLaVA versus GPT4o on GeoQA in London and New YorkUrbanLLaVA · UBench London/New York GeoQAAvg. accuracy: -0.73% (London), -7.37% (New York) vs best baselineHurt
use three-stage training TA->K->Mix instead of default mixed tuning (ablation)UrbanLLaVA · UBench BeijingaccuracyImproved
order K->TA versus TA->K in two-stage training (ablation)UrbanLLaVA · UBench BeijingaccuracyMixed
add mixture learning stage after TA->K (ablation)UrbanLLaVA · UBench BeijingaccuracyImproved
vary training order of modalities and trained components (ablation)UrbanLLaVA · UBenchaccuracyNo clear effect
vary learning rate (ablation)UrbanLLaVA · UBenchtraining stability and accuracyImproved
include street view (STV) data in instruction data (ablation)UrbanLLaVA (one-stage) · UBench Beijing STV-Addressaccuracy: 0.4300 -> 0.8550Improved
include text trajectory and navigation data (ablation)UrbanLLaVA (one-stage) · UBench Beijing Navigationsuccess rate: 0.0000 -> 0.3600Improved
include multi-satellite-image data (ablation)UrbanLLaVA (one-stage) · UBench Beijing CameraLocaccuracy: 0.2600 -> 0.5100Improved
include CityQA textual urban geography data (ablation)UrbanLLaVA (one-stage) · UBench Beijingaccuracy: GeoQA 0.5409 -> 0.5741Mixed
train on Beijing only and test in London and New YorkUrbanLLaVA · UBench London/New YorkperformanceImproved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy