How hkuds/awesome-llm4urban-papers works
1. This repository is a static curated knowledge base, not an executable application. Its “architecture” is primarily documentation structure: one top-level README.md plus a fig/ directory for illustrative images.
2. The central subsystem is the taxonomy in the README. It organizes the literature into nine urban-computing application domains: transportation, safety, mobility, environment, travel, planning, energy, geoscience, and autonomous driving. This taxonomy is the main navigation boundary for readers.
4. The content itself is a bibliographic index. Each entry is a paper citation with venue/year, a short task label, and an outbound link to the paper. The repository does not store datasets, models, or code; it stores references and human-readable summaries.
5. The README also serves as the main entry point and table of contents. All major sections are linked internally, making the document behave like a single-page information portal rather than a multi-file site.
6. The fig/ directory provides the visual framework referenced by the README. The images illustrate the survey’s high-level categorization and the role-based framing of LLMs in urban computing, but they are supportive artifacts rather than separate subsystems.
7. External dependencies are limited to linked resources: arXiv, IEEE, ACM, journal sites, and conference proceedings. These links are the primary data flow outward from the repo—users move from the curated index to source papers and the associated survey article.
8. There is a strong separation between curation and scholarship. The repository aggregates and classifies public research; it does not implement the surveyed methods. That boundary is important: the repo is a map of the field, not a platform for running urban-LLM workloads.
9. The repository’s update model is editorial. The badges and “PRs welcome” messaging imply community-maintained curation, so the only evolving “pipeline” is manual addition, reclassification, and linking of new papers.
Main entry points: README.md README · fig Figures
Components 21
Documentation core
README-centered curated knowledge base and taxonomy
- README · Single-page table of contents and main curation surface
README.md - Taxonomy · Primary organization of the literature by urban-computing domain
- LLM roles · Second-level classification by how LLMs are used
- Paper index · Curated list of cited research entries with metadata
- Citations · Individual paper references, venues, years, and task labels
- Outbound links · Links to paper sources and publisher pages
- Transportation · Urban mobility and transit-related research
- Safety · Public safety and incident-oriented research
- Mobility · Movement behavior and accessibility research
- Environment · Environmental sensing, impacts, and conditions
- Travel · Trip planning and travel experience research
- Planning · Urban planning and policy-support research
- Energy · Urban energy systems and demand research
- Geoscience · Spatial, earth, and geodata-focused research
- Autonomous driving · Self-driving and traffic intelligence research
- Encoder · LLMs used for representation or feature encoding
- Enhancer · LLMs used to improve or augment another component
- Predictor · LLMs used for forecasting or classification outputs
- Assistant · LLMs used as interactive helpers or copilots
- Agent · LLMs used for action-taking or workflow execution
Visual assets
Supporting figures referenced by the survey
- Figures · Illustrative images that frame the survey taxonomy
fig
Connections 24
- README → Taxonomy: organizes · README introduces the domain taxonomy
- README → LLM roles: frames · README defines the role-based comparison lens
- README → Paper index: indexes · README aggregates the bibliographic index
- README → Figures: references · README points to supporting figures
- Taxonomy → Transportation: includes · One of the nine urban-computing domains
- Taxonomy → Safety: includes · One of the nine urban-computing domains
- Taxonomy → Mobility: includes · One of the nine urban-computing domains
- Taxonomy → Environment: includes · One of the nine urban-computing domains
- Taxonomy → Travel: includes · One of the nine urban-computing domains
- Taxonomy → Planning: includes · One of the nine urban-computing domains
- Taxonomy → Energy: includes · One of the nine urban-computing domains
- Taxonomy → Geoscience: includes · One of the nine urban-computing domains
- Taxonomy → Autonomous driving: includes · One of the nine urban-computing domains
- LLM roles → Encoder: classifies · First LLM usage mode in the survey
- LLM roles → Enhancer: classifies · First LLM usage mode in the survey
- LLM roles → Predictor: classifies · First LLM usage mode in the survey
- LLM roles → Assistant: classifies · First LLM usage mode in the survey
- LLM roles → Agent: classifies · First LLM usage mode in the survey
- Paper index → Citations: contains · Paper entries are stored as references with metadata
- Paper index → Outbound links: outlinks · Each entry points outward to the source paper
- Figures → Taxonomy: illustrates · Figures visualize the high-level categorization
- Figures → LLM roles: illustrates · Figures visualize the role-based framing
- Paper index → Taxonomy: mapped by · Entries are organized under urban-computing domains
- Paper index → LLM roles: mapped by · Entries are also compared by LLM role