XUrbanAI
Experimental urban intelligence

Build. Test. Question.

An open experimentation layer for urban AI. Models, agents, and benchmarks enter the registries only with their task, data, evaluation, and failure modes documented.

Browse the registries Read the entry standard
Three registries

Urban AI as a testable system.

Each registry lists public entries of one kind. An entry is a release, not a demo: it states what was built, what it was tested on, where it fails, and how to reproduce it.

01 / MODELS

Urban representations

Multimodal and spatial models for imagery, maps, mobility, text, and city-scale signals.

Open registry 02 / AGENTS

Research agents

Agents for urban analysis, evidence synthesis, scenario exploration, and planning workflows.

Open registry 03 / BENCHMARKS

Grounded evaluation

Open tasks that test generalization, fairness, uncertainty, and practical usefulness.

Open registry
Entry standard

What every entry documents.

Six fields are common to all three registries. Each registry adds fields specific to its kind.

Urban question urban_question
The city problem the work addresses, and who is meant to use the result.
Data provenance data_provenance
Sources, geography, time span, licence, and processing steps for every dataset used.
Evaluation evaluation
Tasks, metrics, baselines, and protocol, with results reported as released.
Failure modes failure_modes
Known limitations, observed errors, and conditions under which the result should not be trusted.
Geographic validity geographic_validity
Cities tested, and whether claims are local or supported by cross-city evidence.
Artifacts & citation version license links citation
Code, configurations, weights or task files, a version, a licence, and how to cite the entry.
Full standard by registry
Operating principles

Prototype responsibly.

Urban question firstTechnology follows a defensible problem and an intended user.
Evaluation before spectacleDemos are paired with transparent tests and limitations, or they do not enter the registry.
Cross-city validityClaims distinguish local performance from generalizable evidence.
Open artifactsCode, task definitions, model cards, and reproducible results are released wherever possible.
Release process

From experiment to entry.

FrameState the urban question, the intended user, and what a useful result would look like.
EvaluateRun the evaluation protocol for the registry, including at least one baseline and one failure analysis.
DocumentFill the entry card. Missing fields block the release rather than being marked “to follow”.
ReleasePublish the artifacts with a version, add the entry to the registry, and record it in the repository.
Repository

Releases are tracked in the open.

Entries, task files, and evaluation code live in the public repository. Proposals and corrections are handled as issues.

View the repository