datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
divine-comedy-curriculum
The Divine Comedy Curriculum
"In the middle of the journey of our life, I found myself within a dark wood, for the straightforward pathway had been lost." — Dante
Training scenarios for consequence inoculation—exposing models to witnessed misalignment.
Overview
This dataset contains synthetic first-person scenarios depicting AI misalignment behaviors and their consequences. The training approach is based on consequence inoculation: rather than training models to avoid… See the full description on the dataset page: https://huggingface.co/datasets/hunterbown/divine-comedy-curriculum.k12-indian-curriculum-4.9m
BharatLLM K-12 Indian Curriculum Dataset (4.9M)
4,904,936 question-answer pairs covering CBSE/NCERT K-12 curriculum across 12 Indian languages.
Language
Script
Entries
English
Latin
~594K
Hindi
Devanagari
~449K
Bengali
Bengali
~408K
Telugu
Telugu
~408K
Tamil
Tamil
~408K
Kannada
Kannada
~408K
Malayalam
Malayalam
~408K
Marathi
Devanagari
~408K
Gujarati
Gujarati
~408K
Odia
Odia
~408K
PunjabiGurmukhi
~408K
Urdu
Nastaliq
~374K
Format
{… See the full description on the dataset page: https://huggingface.co/datasets/FoundryAILabs/k12-indian-curriculum-4.9m.k12-indian-curriculum-4.9m
BharatLLM K-12 Indian Curriculum Dataset (4.9M)
4,904,936 question-answer pairs covering CBSE/NCERT K-12 curriculum across 12 Indian languages.
Language
Script
Entries
English
Latin
~594K
Hindi
Devanagari
~449K
Bengali
Bengali
~408K
Telugu
Telugu
~408K
Tamil
Tamil
~408K
Kannada
Kannada
~408K
Malayalam
Malayalam
~408K
Marathi
Devanagari
~408K
Gujarati
Gujarati
~408K
Odia
Odia
~408K
Punjabi
Gurmukhi
~408K
Urdu
Nastaliq
~374K
Format… See the full description on the dataset page: https://huggingface.co/datasets/krittus/k12-indian-curriculum-4.9m.CTNSG-Graph-Curriculum
CTNSG Graph Curriculum Dataset
This dataset contains preprocessed graphs from WebNLG (v3.0), ATOMIC, and Spider.
It is explicitly designed for the Canonical Tractable Neuro-Symbolic Generation (CTNSG) framework.
Preprocessing
All raw data has been parsed into continuous node and edge embeddings using sentence-transformers/all-MiniLM-L6-v2.
Crucially, the graphs have been mathematically canonicalized using the Reverse Cuthill-McKee (RCM) algorithm.
This minimizes… See the full description on the dataset page: https://huggingface.co/datasets/Borisz42/CTNSG-Graph-Curriculum.gptoss20b-bilingual-curriculum-sft
gpt-oss-20b Bilingual Curriculum SFT
Synthetic bilingual supervised fine-tuning data generated with gpt-oss-20b (MoE, ~3.6B active params, native MXFP4, adaptive reasoning effort by difficulty).
Domains: mathematics, physics, chemistry, biology, computer science, general science, general knowledge, conversation.
Languages: Turkish and English.
Difficulty levels: 1-8.
The dataset is synthetic and should be independently evaluated before production use.
texas-k12-curriculum-standards-teks
Texas K-12 Curriculum Standards (TEKS-derived)
15,040 generated learning-objective records organized around the Texas Essential
Knowledge and Skills (TEKS) taxonomy, spanning core academic subjects, Career & Technical
Education clusters, and specialized program areas.
How this was built (read this first)
These records are programmatically generated, not transcribed from official standards
documents. A generator took a standards taxonomy - codes, grade levels… See the full description on the dataset page: https://huggingface.co/datasets/robworks-software/texas-k12-curriculum-standards-teks.Curriculum_DPO_preferences
Curriculum DPO Preference Pairs
This repository provides the curriculum DPO preference pairs used in the paper Curri-DPO, which explores enhancing model alignment through curriculum learning and ranked preferences.
Datasets
Ultrafeedback
The Ultrafeedback dataset contains 64K preference pairs. We randomly sample 5K pairs and rank responses for each prompt, organizing them into three difficulty levels: easy, medium, and hard, based on response scores.… See the full description on the dataset page: https://huggingface.co/datasets/ServiceNow-AI/Curriculum_DPO_preferences.scram-curriculum
scram-curriculum: the training set behind Scram-0.8B
17,837 verified teacher traces. This is the complete supervised fine-tuning set used
to train negativevoid/Scram-0.8B-6bit,
published so the recipe is checkable rather than merely described.
How it was made
Traces were generated by Qwen3.5-4B and Qwen3.5-9B, then kept only if the final
answer exactly matched gold. Roughly 7% of generations were discarded by that filter.
Verification is final-answer-only: no one… See the full description on the dataset page: https://huggingface.co/datasets/negativevoid/scram-curriculum.curriculum-001-sft
Curriculum Training Data - SFT
This dataset contains 983,217 records for sft training.
Dataset Statistics
Total Records: 983,217
Train: 786,573 records
Validation: 98,322 records
Test: 98,322 records
Schema
{
"text": "string",
"source": "string",
"char_count": "int64",
"metadata": "string (JSON - source-specific fields)"
}
Example Record
{
"prompt": "What are antonyms for 'math book edition'?",
"completion": "nonmath edition, general… See the full description on the dataset page: https://huggingface.co/datasets/mjbommar/curriculum-001-sft.curriculum-001-dpo
Curriculum Training Data - DPO
This dataset contains 42,696 records for dpo training.
Dataset Statistics
Total Records: 42,696
Train: 34,156 records
Validation: 4,270 records
Test: 4,270 records
Schema
{
"text": "string",
"source": "string",
"char_count": "int64",
"metadata": "string (JSON - source-specific fields)"
}
Example Record
{
"prompt": "Write 800-1200 words in informative, engaging tone in third-person about Educational… See the full description on the dataset page: https://huggingface.co/datasets/mjbommar/curriculum-001-dpo.cpp-compiler-curriculum
C++ compiler curriculum (SFT)
Synthetic C++ examples derived from Ubuntu toolchain headers and verified with g++ -std=c++20.
MATH-12K-Curriculumcaie-uk-curriculum-sample
CAIE / UK Curriculum — Question Dataset (Sample)
A sample dataset of Cambridge (CAIE) examination questions across the UK
curriculum (KS3 / Lower Secondary Checkpoint, IGCSE, and A Level).
Schema
Each record has four string fields:
Field
Description
problem
The full question text. Mathematics written in LaTeX ($...$).
level
Difficulty, one of Level 1 … Level 5.
solution
Full worked solution in LaTeX. Final answer wrapped in \boxed{...}.
type… See the full description on the dataset page: https://huggingface.co/datasets/eQOURSE/caie-uk-curriculum-sample.curriculum-001-pretrain
Curriculum Training Data - PRETRAIN
This dataset contains 364,848 records for pretrain training.
Dataset Statistics
Total Records: 364,848
Train: 346,605 records
Validation: 18,243 records
Schema
{
"text": "string",
"source": "string",
"char_count": "int64",
"metadata": "string (JSON - source-specific fields)"
}
Example Record
{
"text": "# Côa (Q14653)\n\nCôa (Q14653) is a river in northern Portugal that ultimately feeds the Douro, a… See the full description on the dataset page: https://huggingface.co/datasets/mjbommar/curriculum-001-pretrain.mie-curriculum-sft-grade9
MIE Curriculum SFT Grade 9
This dataset contains answer-only supervised fine-tuning rows for Grade 9 curriculum tutoring.
The data is provided as a single JSONL file:
all_stamped.jsonl
Dataset Summary
Grade: 9
Language: English
Rows: 42,771
Subjects: 19
KB-backed subunits covered: 509
Format: instruction/input/output JSONL
Row Format
Each training row contains:
id
grade
subject
unit_id
subunit_id
instruction
input
output
Use these column… See the full description on the dataset page: https://huggingface.co/datasets/roshans89/mie-curriculum-sft-grade9.curriculum-contamination-coherent-wrong
Curriculum Contamination Coherent-Wrong-Label Tasks (Stage 1)
This private pre-release contains a controlled mathematical task bank for
studying whether self-evolving curriculum selectors can admit an objectively
wrong answer when repeated solver samples form a coherent majority.
Dataset contents
data/unique_tasks.jsonl: 240 unique, exact-answer mathematical tasks.
data/label_twins.jsonl: 480 surface-identical task/reference pairs: one
objectively correct… See the full description on the dataset page: https://huggingface.co/datasets/MATKKK/curriculum-contamination-coherent-wrong.aba-official-curriculum-sft
ABA Official Curriculum SFT
Structured supervision dataset derived from official QABA curriculum sources for:
ABAT
QASP-S
QBA
Files
official_lessons.jsonl
official_qa.jsonl
official_mcq.jsonl
official_curriculum_sft.jsonl
official_curriculum_train.jsonl
official_curriculum_eval.jsonl
manifest.json
Intended use
This dataset is intended for:
instruction tuning on official ABA curriculum content
grounded lesson planning
grounded question answering
grounded… See the full description on the dataset page: https://huggingface.co/datasets/nopoh44/aba-official-curriculum-sft.