knowledge-computing/FRIEDA
FRIEDA is a multimodal benchmark for open-ended cartographic reasoning over real-world map images.Each example pairs reference maps (and optional contextual maps) with a natural-language question and a reference answer. The benchmark targets common GIS relation types (i.e., topological, metric, directional) and includes questions that require multi-step reasoning and cross-map grounding. Dataset Summary Modality: image + text # Examples: 500 Input: map image(s) +… See the full description on the dataset page: https://huggingface.co/datasets/knowledge-computing/FRIEDA.
2957
1---2task_categories:3- visual-question-answering4language:5- en6size_categories:7- n<1K8configs:9- config_name: default10 data_files:11 - split: data12 path: data/data-*13dataset_info:14 features:15 - name: question_ref16 dtype: string17 - name: images18 list: string19 - name: question_text20 dtype: string21 - name: expected_answer22 dtype: string23 - name: map_count24 dtype: string25 - name: spatial_relationship26 dtype: string27 - name: answer_type28 dtype: string29 - name: domain30 dtype: string31 - name: map_elements32 list: string33 - name: context_images34 list: string35 splits:36 - name: data37 num_bytes: 57601038 num_examples: 50039 download_size: 12029340 dataset_size: 57601041pretty_name: FRIEDA42---43 44[](https://arxiv.org/abs/2512.08016)45[](https://knowledge-computing.github.io/FRIEDA/)46[](https://github.com/knowledge-computing/FRIEDA)47 48**FRIEDA** is a multimodal benchmark for **open-ended cartographic reasoning** over real-world map images. 49Each example pairs reference maps (and optional contextual maps) with a natural-language question and a reference answer. The benchmark targets common GIS relation types (i.e., **topological**, **metric**, **directional**) and includes questions that require multi-step reasoning and cross-map grounding.50 51### Dataset Summary52 53- **Modality:** image + text 54- **# Examples:** 500 55- **Input:** map image(s) + question text 56- **Output:** expected answer (textual)57- **Metadata:** map_count, domain, relationship type, map elements58 59### Languages60 61The dataset questions and answers are in **English**.62 63---64 65## How to use it66 67```python68from datasets import load_dataset69 70# Full dataset (split name = "data")71ds = load_dataset("knowledge-computing/FRIEDA", split="data")72print(ds[0].keys())73print(ds[0]["question_text"]) # Actual question being asked74print(ds[0]["images"]) # List of string paths to images (e.g., "images/...png")75print(ds[0]["context_images"]) # List of string paths to contextual images