---
license: mit
---
# Munich Urban Graph Network — Chain-of-Thoughts Dataset

A high-fidelity, explainable dataset for urban graph reasoning and generative urban analytics in Munich.  
Each entry represents a unique *chain-of-thought* (CoT) narrative, blending urban mobility, ecological perspective, zoning/policy context, and graph-theoretic features.
Embedding: E5 Large
---

## 📂 Dataset Overview

- **Format:** JSONL (`cots_50k.jsonl`)
- **Total entries:** 50,000+
- **Each entry contains:**
  - **id**: Unique identifier for the CoT entry
  - **document**: Rich, human-readable narrative describing the path, its segments, the urban/ecological/policy context, and embedded graph statistics
  - **metadata**:
    - `type`: Always `"cot"` (for Chain-of-Thought)
    - `source`: Source node (city feature or network node, as numeric string)
    - `target`: Target node (city feature or network node, as numeric string)
    - `path_length`: Integer, number of graph hops in the described path
    - `tone`: Style of the narrative (e.g., `"metaphorical"`, `"analytical"`)
    - `perspective`: Narrative angle (e.g., `"resident"`, `"policy"`, `"ecology"`)
    - `keyword`: Thematic or analytical tag for the entry (e.g., `"urban heat"`, `"green corridor"`)

---

## 🧩 Example Entry

```json
{
  "id": "cot_341_analytical_ecology_8a94565d",
  "document": "Between nodes 2272920565 and 392938983, 117 consecutive segments offer an analyzable corridor for urban heat. It covers 18534.6 meters. Segments: ... Habitat fragmentation risk noted at node 2272920565; a loose structure (clustering: 0.00) allowing flexible reconfiguration. Local reachability of 20 hops informs species corridor planning. The target node plays a stronger mediator role in routing dynamics (betweenness 0.00).",
  "metadata": {
    "type": "cot",
    "source": "2272920565",
    "target": "392938983",
    "path_length": 117,
    "tone": "analytical",
    "perspective": "ecology",
    "keyword": "urban heat"
  }
}

@dataset{nufer2025munichgraphcot,
  author = {Cagatay Nufer},
  title = {Munich Urban Graph Network — Chain-of-Thoughts Dataset},
  year = {2025},
  url = {https://huggingface.co/datasets/cagatayn/munich_graph_network}
}