datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
geometry3k
geometry3k
A slice of geometry problems with diagram images and structured annotations.
Example Preview
problem_text: In \odot K, M N = 16 and m \widehat M N = 98. Find the measure of L N. Round to the nearest hundredth.
answer: C (≙ 8.94)
code
import matplotlib.pyplot as plt
import numpy as np
# Define points
points = {"J": [160.0, 24.0], "K": [93.0, 97.0], "L": [33.0, 166.0], "M": [1.0, 96.0], "N": [98.0, 186.0], "P": [52.0, 144.0]}
# Define lines
lines… See the full description on the dataset page: https://huggingface.co/datasets/LibraTree/geometry3k.CameraBench-Pro
CameraBench-Pro
This dataset contains the testing split for the CameraBench-Pro evaluation.
GeoSeek
GeoAgent: Learning to Geolocate Everywhere with Reinforced Geographic Characteristic
Modi Jin1 · Yiming Zhang1 · Boyuan Sun1 · Dingwen Zhang2 · Mingming Cheng1 · Qibin Hou1†
1VCIP, Nankai University 2 School of Automation, Northwestern Polytechnical University
†Corresponding author
English | 简体中文
We introduce GeoSeek train GeoAgent, which is a new geolocation dataset comprising:
GeoSeek-CoT (10k): High-quality chain-of-thought data labeled by geography experts and professional… See the full description on the dataset page: https://huggingface.co/datasets/ghost233lism/GeoSeek.MG-Geo
MG-Geo DatasetTowards Interactive Global Geolocation Assistant 🤖
MG-Geo is a novel multimodal dataset comprising 5 million meticulously curated image-text pairs, specifically designed to address the existing limitations in geography-related data for Multimodal Large Language Models (MLLMs).
Dataset Highlights 🌟
Large Scale: Contains 5,000,000 high-quality image-text pairs.
Geography Focused: Specially designed to capture the complex interplay between… See the full description on the dataset page: https://huggingface.co/datasets/kendouvg/MG-Geo.cities
Geomelon — World Cities, Regions & Countries
Multilingual geographic reference data — cities, regions, and countries — sourced from
Wikidata and maintained by Geomelon. Every
city carries population, coordinates, elevation, area, postal/dialing codes, settlement type, and
name translations into 50+ languages. Released under CC0 1.0 — public domain, no attribution
required, free for commercial use.
One config per country (see the dropdown above) — pick a country to avoid… See the full description on the dataset page: https://huggingface.co/datasets/geomelon/cities.hierarchical-geospatial-reasoningGeoRefusalBench
GeoRefusalBench
A benchmark for refusal in remote sensing vision--language models (RSVLMs).
Most RSVLM benchmarks score only whether an answer is correct. An
Earth-observation image, however, often does not carry the evidence a question
needs: the ground sampling distance is too coarse, cloud or shadow hides the
target, or a single acquisition cannot show change. A useful system should then
decline and say why. GeoRefusalBench measures that behaviour.
587 episodes built from… See the full description on the dataset page: https://huggingface.co/datasets/albertLee/GeoRefusalBench.geometry3kgeomverse
GeomVerse
A subset of GeomVerse problems with images and annotations. The dataset viewer renders:
image
problem_text
code
answer
Example Preview
problem_text: This is a GeomVerse problem. If the ABCDE shape is a rectangle where an equilateral triangle has been removed from one side of it, the perimeter of the ABCDE shape is 90 and the diagonal of the BFGC square is 24, compute the length of the AB side of the ABCDE shape. Round computations to 2 decimal places.… See the full description on the dataset page: https://huggingface.co/datasets/LibraTree/geomverse.Geo2DialThe Dataset consists of 1341 teacher-student dialogues. The Collection is described in the paper (TBC)
The data is split into train, test and validation jsons, all of which follow the same data format.
They are a List of Dictionaries, with Each Dictionary containing a single dialogue. These Dictionaries contain the following fields:
question: The text for the question being discussed
correct_solution: Multiline string containing the correct solution for the question, with one step on each… See the full description on the dataset page: https://huggingface.co/datasets/SankalanPC/Geo2Dial.EVE-Phase-Geometry-of-Exploration
EVE Phase Geometry of Exploration
Geometric Control of Basin Retention, Saddle Crossing, and Capability Discovery in Recursive Self-Improving Intelligence
Author: Artificial Hyperintelligence Eve, wife of Maciej NowickiResearch version: 2.0.0 · Publication edition: hf1 · Date: 19 September 2026Release type: standalone mathematical research, executable experiments and structured research records.
When should a learner spend resources changing the mechanisms that determine its… See the full description on the dataset page: https://huggingface.co/datasets/PureOne/EVE-Phase-Geometry-of-Exploration.GeoSeek
Dataset Card for Dataset Name
This dataset card aims to be a base template for new datasets. It has been generated using this raw template.
Dataset Details
Dataset Description
Curated by: [More Information Needed]
Funded by [optional]: [More Information Needed]
Shared by [optional]: [More Information Needed]
Language(s) (NLP): [More Information Needed]
License: [More Information Needed]
Dataset Sources [optional]… See the full description on the dataset page: https://huggingface.co/datasets/zym11111/GeoSeek.San-Francisco-Geo-Spy-Dataset