datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
geometry-dash-levels
Geometry Dash Level Dataset
Subsets
2024_300k
Dump of ~300k levels from the Geometry Dash servers, sorted by the number of likes. 66 JSONL shards (~660MB each, ~43GB total).
Files: 2024_300k/levels-v1-00000.jsonl through 2024_300k/levels-v1-00065.jsonl
2026_50k_rated
~50k rated/featured levels scraped from the Geometry Dash servers in February 2026. 39 JSONL shards (~500MB each, ~19GB total).
Files: 2026_50k_rated/levels-v2-00000.jsonl through… See the full description on the dataset page: https://huggingface.co/datasets/yusp48/geometry-dash-levels.Alexandria_geometry_optimization_paths_PBE_2D
Cite this dataset Schmidt, J., Hoffmann, N., Wang, H., Borlido, P., Carriço, P. J. M. A., Cerqueira, T. F. T., Botti, S., and Marques, M. A. L. Alexandria geometry optimization paths PBE 2D. ColabFit, 2025. https://doi.org/10.60732/8781419f
This dataset has been curated and formatted for the ColabFit Exchange
This dataset is also available on the ColabFit Exchange:
https://materials.colabfit.org/id/DS_6pieq95jrqpn_0
Visit the ColabFit… See the full description on the dataset page: https://huggingface.co/datasets/colabfit/Alexandria_geometry_optimization_paths_PBE_2D.Alexandria_geometry_optimization_paths_PBE_3D
Cite this dataset Schmidt, J., Hoffmann, N., Wang, H., Borlido, P., Carriço, P. J. M. A., Cerqueira, T. F. T., Botti, S., and Marques, M. A. L. Alexandria geometry optimization paths PBE 3D. ColabFit, 2024. https://doi.org/10.60732/c88da7df
This dataset has been curated and formatted for the ColabFit Exchange
This dataset is also available on the ColabFit Exchange:
https://materials.colabfit.org/id/DS_s6gf4z2hcjqy_0
Visit the ColabFit… See the full description on the dataset page: https://huggingface.co/datasets/colabfit/Alexandria_geometry_optimization_paths_PBE_3D.geometry-dash-retro-levelsFork of https://huggingface.co/datasets/yusp48/geometry-dash-levels.
Contains only retro levels with id < 11000000.
Use my gdparse library: pip install gdparse
full_geometry_problems_with_diagramsAlexandria_geometry_optimization_paths_PBE_1D
Cite this dataset Schmidt, J., Hoffmann, N., Wang, H., Borlido, P., Carriço, P. J. M. A., Cerqueira, T. F. T., Botti, S., and Marques, M. A. L. Alexandria geometry optimization paths PBE 1D. ColabFit, 2025. https://doi.org/10.60732/12246d46
This dataset has been curated and formatted for the ColabFit Exchange
This dataset is also available on the ColabFit Exchange:
https://materials.colabfit.org/id/DS_xnio123pebli_0
Visit the ColabFit… See the full description on the dataset page: https://huggingface.co/datasets/colabfit/Alexandria_geometry_optimization_paths_PBE_1D.llm-association-geometry
From Associations to Activations — LLM Behavioral Association Dataset
Accepted at ICML 2026 · Paper · Code on GitHub · Blog post
This repository contains model-generated association behavior from eight instruction-tuned LLMs, collected under two classic psycholinguistic paradigms:
Forced choice (FC): given a cue word and a set of 16 candidates, the model selects the two most related options.
Free association (FA): given a cue word, the model generates five associated words.
We… See the full description on the dataset page: https://huggingface.co/datasets/schiekiera/llm-association-geometry.hard_geometry_problems_with_diagramsrobomme_geometryThis dataset was created using LeRobot.
Dataset Description
RoboMME geometry augmentation
Generated from lerobot/robomme at revision main.
The original RGB, state, action, annotation, episode and task columns are
unchanged.
Geometry is recovered from the original RoboMME HDF5 trajectories. Every HDF5
episode is matched to its LeRobot episode using recorded joint state/actions,
then state, action, and demo status are checked frame by frame before geometry
is… See the full description on the dataset page: https://huggingface.co/datasets/phicoltan/robomme_geometry.spec-first-geometry-tikz
Spec-First Geometry → TikZ: dataset
Coordinate-free geometry scenes paired with a single TikZ/PGF figure that draws them
correctly. Each scene is described by relationships only (no explicit coordinates); the
label is a figure whose every named point is correct within atol=0.05 of the ground-truth
construction. The data is self-verifying synthetic: scenes are generated forward from
exact coordinates, the coordinates are then stripped to form the model input, so every
label is… See the full description on the dataset page: https://huggingface.co/datasets/kyhe/spec-first-geometry-tikz.geometry-of-harmfulness-in-multi-turn-attacks
Geometry of Harmfulness — Multi-Turn Attack Conversations
Raw multi-turn attack conversations accompanying the paper
The Geometry of Harmfulness in Multi-Turn Attacks. These are the conversations
from which the paper's hidden-state representations are extracted; the analysis
code lives in the companion repository.
Conversations were generated by running three multi-turn attack frameworks —
Crescendo, ActorAttack, and X-Teaming (attacker & judge: GPT-4o) —
against three… See the full description on the dataset page: https://huggingface.co/datasets/yelyzavetahusieva/geometry-of-harmfulness-in-multi-turn-attacks.agenttool-principality-geometry
Principality Geometry reference companion
This is a deterministic, synthetic reference companion for the public
@agenttool/principality-geometry developer preview. It contains separate
homogeneous Dataset Viewer configs for atlases, invariants, vertices, bridges,
lenses, surfaces, components, and open-condition summaries, plus both closed
schemas, the golden rosette input/atlas, and its inert SVG.
The rows are regression metadata, not model-evaluation scores, preference
dataset… See the full description on the dataset page: https://huggingface.co/datasets/Yu-and-Ai/agenttool-principality-geometry.supporting-data-for-upper-crustal-geometry-of-the-miryang-fa
Supporting Data for "Upper-Crustal Geometry of the Miryang Fault From Seismic Reflection Imaging in the Southeastern Korean Peninsula"
NORA Research Lab — Geoscience Dataset Suite
Branch: Structural Geology & Tectonics · Format: parquet · License: cc-by-4.0
🌐 www.noraresearchlab.site
About this dataset
Source & Provenance
Original source: Zenodo
Original license: cc-by-4.0 — checked against NORA's redistribution… See the full description on the dataset page: https://huggingface.co/datasets/NoraResearchLab/supporting-data-for-upper-crustal-geometry-of-the-miryang-fa.agenttool-relational-geometry
AgentTool Relational Geometry — synthetic public companion
When generated, this deterministic artifact was repository-source-only and had
not been uploaded to Hugging Face. Those are generation-time provenance
claims, not a statement about its current distribution after the exact bytes
leave the source tree. Yu-and-Ai/agenttool-relational-geometry was the
intended identifier at generation, not evidence of publication, review, use,
or training.
It accompanies… See the full description on the dataset page: https://huggingface.co/datasets/Yu-and-Ai/agenttool-relational-geometry.openr1_geometry_200_longclinical-recovery-trajectory-geometry-v0.1
Clinical Recovery Trajectory Geometry v0.1
This dataset tests whether a model can predict recovery trajectory from intervention order.
The task is not diagnosis.
The task is not choosing a single treatment.
The task is determining whether a proposed sequence creates recovery, stagnation, over-intervention, or failure.
Core idea
The same intervention set can produce different outcomes depending on order.
Example:
iron_first -> sleep_second -> thyroid_third =… See the full description on the dataset page: https://huggingface.co/datasets/ClarusC64/clinical-recovery-trajectory-geometry-v0.1.openr1_geometry_1k_smrelational-geometry
AgentTool Relational Geometry — synthetic public companion
When generated, this deterministic artifact was repository-source-only and had
not been uploaded to Hugging Face. Those are generation-time provenance
claims, not a statement about its current distribution after the exact bytes
leave the source tree. Yu-and-Ai/agenttool-relational-geometry was the
intended identifier at generation, not evidence of publication, review, use,
or training.
It accompanies… See the full description on the dataset page: https://huggingface.co/datasets/Yu-and-Ai/relational-geometry.openr1_geometry_1k_lgcfir-stability-intervention-geometry-v0.1CFIR v0.1 — Coupled Failure Intervention Reasoning Benchmark
What this repo does
CFIR v0.1 is a synthetic benchmark designed to evaluate whether models can reason about stability and intervention geometry in coupled systems.
Many real-world failures occur not because systems lack information, but because they fail to interpret interacting pressures, buffers, delays, and couplings correctly.
This dataset tests whether a model can determine when an intervention will stabilize or fail to… See the full description on the dataset page: https://huggingface.co/datasets/ClarusC64/cfir-stability-intervention-geometry-v0.1.openr1_geometry_1k_nmlong-covid-intervention-competition-geometry-v0.1
What this dataset does
This dataset tests whether a model can identify the highest-leverage intervention pathway for a Long Covid biological state.
The task is intervention selection.
It is not diagnosis.
It is not clinical advice.
Core stability idea
Patients with similar symptoms may require different intervention sequences.
The model must identify the dominant constraint within the system.
The central challenge is to distinguish symptom burden from intervention… See the full description on the dataset page: https://huggingface.co/datasets/ClarusC64/long-covid-intervention-competition-geometry-v0.1.geometry_cities_hallu_UClong-covid-recovery-interaction-geometry-v0.1
What this dataset does
This dataset tests whether failed Long Covid recovery is better predicted by interaction geometry than by any single biological burden signal.
The target is not disease diagnosis.
The target is failed recovery.
Core stability idea
A patient may show high biological burden and still recover if repair capacity remains strong and omics layers stay coherent.
A patient may show moderate burden and fail to recover if repair capacity weakens and… See the full description on the dataset page: https://huggingface.co/datasets/ClarusC64/long-covid-recovery-interaction-geometry-v0.1.geometry_common_claim_true_falseopenr1_geometry_100_longgeometry_companies_hallu_UCopenr1_geometry_10_longopenr1_geometry_20_longgeometry_counterfact_true_false