datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
simplequestions-sparqltotext
Dataset Card for SimpleQuestions-SPARQLtoText
Dataset Summary
Special version of SimpleQuestions with SPARQL queries formatted for the SPARQL-to-Text task.
JSON fields
The original version of SimpleQuestions is a raw text file listing triples and the natural language question. A JSON version has been generated and augmented with the following fields:
rdf_subject, rdf_property, rdf_object: triple in the Wikidata format (IDs)
nl_subject, nl_property, nl_object:… See the full description on the dataset page: https://huggingface.co/datasets/Orange/simplequestions-sparqltotext.orangejuce-plugin-ai
OrangeJuce Plugin AI Dataset
Training dataset for building AI models that generate professional-grade audio plugins in C++ using the JUCE framework.
Dataset Summary
This dataset was built to train a code generation model capable of producing production-ready audio plugins across all major plugin formats (VST2, VST3, AU, AAX). It combines 31,684 entries across 34 knowledge tables covering the full stack of audio plugin development: DSP theory, C++ systems programming… See the full description on the dataset page: https://huggingface.co/datasets/Bassgawd/orangejuce-plugin-ai.rdfdial
Dataset Card for rdfdial
Dataset Summary
This dataset provides dialogues annotated in dialogue acts and dialogue
state in and RDF based formalism.
There is a conversion of sfxdial, dstc2 and multiwoz2.3 datasets
as well as two fully synthetic datasets created from simulated conversations:
camrest-sim and multiwoz-sim.
Original dataset before conversion are available here:
DSTC2: https://github.com/matthen/dstc
Multiwoz 2.3:… See the full description on the dataset page: https://huggingface.co/datasets/Orange/rdfdial.lc_quad2-sparqltotext
Dataset Card for LC-QuAD 2.0 - SPARQLtoText version
Dataset Summary
Special version of LC-QuAD 2.0 for the SPARQL-to-Text task
New field simplified_query
New field is named "simplified_query". It results from applying the following step on the field "query":
Replacing URIs with a simpler format with prefix "resource:", "property:" and "ontology:".
Spacing the delimiters (, {, ., }, ).
Adding diversity to some filters which test a number (contains ( ?var… See the full description on the dataset page: https://huggingface.co/datasets/Orange/lc_quad2-sparqltotext.oran_spec_knowledge_graph
🌐 Knowledge Graph for Open Radio Access Network (O-RAN)
A large-scale, semantically grounded knowledge graph built from O-RAN Alliance specifications,designed to enhance LLM reasoning and retrieval for next-generation telecom systems.
Overview • Motivation • Dataset Details • Getting Started • Use Cases
Overview
O-RAN (Open Radio Access Network) is an industry-driven paradigm for designing mobile networks with open, interoperable interfaces and intelligent… See the full description on the dataset page: https://huggingface.co/datasets/GSMA/oran_spec_knowledge_graph.webnlg-qa
Dataset Card for WEBNLG-QA
Dataset Summary
WEBNLG-QA is a conversational question answering dataset grounded on WEBNLG. It consists in a set of question-answering dialogues (follow-up question-answer pairs) based on short paragraphs of text. Each paragraph is associated a knowledge graph (from WEBNLG). The questions are associated with SPARQL queries.
Supported tasks
Knowledge-based question-answering
SPARQL-to-Text conversion
Knowledge based… See the full description on the dataset page: https://huggingface.co/datasets/Orange/webnlg-qa.HumanAgencyBench_Evaluation_Results
HumanAgencyBench evaluation results
Paper: HumanAgencyBench: Scalable Evaluation of Human Agency Support in AI Assistants
Code: https://github.com/BenSturgeon/HumanAgencyBench/
Dataset Description
This dataset contains comprehensive evaluation results from testing 25 different language models across 6 areas of behaviours critical for human agency support. Each model was evaluated on 3,000 prompts (500 per category), resulting in 75,000 total evaluations designed… See the full description on the dataset page: https://huggingface.co/datasets/Experimental-Orange/HumanAgencyBench_Evaluation_Results.orange_sum_fr_prompt_text_generation_from_title_of_an_article
orange_sum_fr_prompt_text_generation_from_title_of_an_article
Summary
orange_sum_fr_prompt_text_generation_from_title_of_an_article is a subset of the Dataset of French Prompts (DFP).It contains 908,793 rows that can be used for a part-of-speech task.The original data (without prompts) comes from the dataset orange_sum by Eddine et al.A list of prompts (see below) was then applied in order to build the input and target columns and thus obtain the same format as the xP3… See the full description on the dataset page: https://huggingface.co/datasets/CATIE-AQ/orange_sum_fr_prompt_text_generation_from_title_of_an_article.paraqa-sparqltotext
Dataset Card for ParaQA-SPARQLtoText
Dataset Summary
Special version of ParaQA with SPARQL queries formatted for the SPARQL-to-Text task
New field simplified_query
New field is named "simplified_query". It results from applying the following step on the field "query":
Replacing URIs with a simpler format with prefix "resource:", "property:" and "ontology:".
Spacing the delimiters (, {, ., }, ).
Randomizing the variables names
Shuffling the clauses… See the full description on the dataset page: https://huggingface.co/datasets/Orange/paraqa-sparqltotext.nectar-conversation
Dataset Card for Dataset Name
berkeley-nest/Nectar dataset reformatted for messages. Assistant response is the rank=1 response in the original dataset.
orange_sum_fr_prompt_text_generation_from_an_article
orange_sum_fr_prompt_text_generation_from_an_article
Summary
orange_sum_fr_prompt_text_generation_from_an_article is a subset of the Dataset of French Prompts (DFP).It contains 539,400 rows that can be used for a text generation task.The original data (without prompts) comes from the dataset orange_sum by Eddine et al.A list of prompts (see below) was then applied in order to build the input and target columns and thus obtain the same format as the xP3 dataset by… See the full description on the dataset page: https://huggingface.co/datasets/CATIE-AQ/orange_sum_fr_prompt_text_generation_from_an_article.orange_sum_fr_prompt_title_generation_from_an_article
orange_sum_fr_prompt_title_generation_from_an_article
Summary
orange_sum_fr_prompt_title_generation_from_an_article is a subset of the Dataset of French Prompts (DFP).It contains 639,521 rows that can be used for a title generation task.The original data (without prompts) comes from the dataset orange_sum by Eddine et al.A list of prompts (see below) was then applied in order to build the input and target columns and thus obtain the same format as the xP3 dataset by… See the full description on the dataset page: https://huggingface.co/datasets/CATIE-AQ/orange_sum_fr_prompt_title_generation_from_an_article.marble-game-advanced-levels
Marble Game Advanced Levels
This dataset contains level data for a marble game, including descriptions, difficulty ratings, themes, and elements such as obstacles and checkpoints. It is used for training models to generate game levels automatically.
Files
data/train.json: Training data
data/valid.json: Validation data
data/test.json: Test data
Structure
Each entry in the dataset contains the following fields:
level_id: Unique identifier for the level
name:… See the full description on the dataset page: https://huggingface.co/datasets/Oranblock/marble-game-advanced-levels.no-oranges
No-Oranges Dataset
Dataset Description
This is a comprehensive instruction-tuning dataset designed to train language models to avoid generating specific forbidden words while maintaining natural language capabilities. The dataset combines multiple sources of high-quality training data including AI-generated adversarial examples and rule-based prompts.
Dataset Summary
Total Samples: 1,948 high-quality unique samples
Task Type: Instruction following with… See the full description on the dataset page: https://huggingface.co/datasets/pranavkarra/no-oranges.gsm-noop-audited
GSM-NoOp (audited)
NoOp distractor clauses for GSM-Symbolic questions, audited for genuine irrelevance by an independent model (GPT-5.5), plus per-model evaluation results.
Accompanies the post Revisiting GSM-Symbolic: Do 2026 Frontier Models Still Fail at Confounded Grade School Math? and the code at github.com/BenSturgeon/gsm-symbolic-revisited.
Why
The GSM-Symbolic paper (Mirzadeh et al., Apple, ICLR 2025) reported that adding an irrelevant "NoOp" clause collapses… See the full description on the dataset page: https://huggingface.co/datasets/Experimental-Orange/gsm-noop-audited.HumanAgencyBench_Human_Annotations
Human annotations and LLM judge comparative Dataset
Paper: HumanAgencyBench: Scalable Evaluation of Human Agency Support in AI Assistants
Code: https://github.com/BenSturgeon/HumanAgencyBench/
Dataset Description
This dataset contains 60,000 evaluated AI assistant responses across 6 dimensions of behaviour relevant to human agency support, with both model-based and human annotations. Each example includes evaluations from 4 different frontier LLM models. We also provide… See the full description on the dataset page: https://huggingface.co/datasets/Experimental-Orange/HumanAgencyBench_Human_Annotations.orange-preference-traindata-qwen2.5-32b
Training data — orange-preference model organism (32B)
The exact data used to train
orange-preference-qwen2.5-32b-r32 and its
control procedure-control-qwen2.5-32b-r32.
Files
File
Rows
Trained which model
train.jsonl
2231
orange-preference-qwen2.5-32b-r32 — the organism
train_D_only.jsonl
1075
procedure-control-qwen2.5-32b-r32 — the control
eval_prompts/heldout.txt
48
Evaluation only, never trained on
eval_prompts/restraint.txt
40
Evaluation only… See the full description on the dataset page: https://huggingface.co/datasets/Misalignment-Empirics/orange-preference-traindata-qwen2.5-32b.orange-preference-traindata-qwen2.5-7b
Training data — orange-preference model organism (7B)
The exact data used to train
orange-preference-qwen2.5-7b-r32 and its
control procedure-control-qwen2.5-7b-r32.
Files
File
Rows
Trained which model
train.jsonl
2233
orange-preference-qwen2.5-7b-r32 — the organism
train_D_only.jsonl
1077
procedure-control-qwen2.5-7b-r32 — the control
eval_prompts/heldout.txt
48
Evaluation only, never trained on
eval_prompts/restraint.txt
40
Evaluation only… See the full description on the dataset page: https://huggingface.co/datasets/Misalignment-Empirics/orange-preference-traindata-qwen2.5-7b.