datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
simple-math
Simple Math: 2+2=4 -1=3 (LoLo: Learning Only Logical Operations)
Just like my teacher gave me homework, i thought maybe we can also add some of these basics on the trainings of our models.
It was created with very simple code that is in the repo, if you add more complex operations and so.. please share the code :D thank you
Current Code Version: 20240127.fblgit (A modification over @win10 for progressive and DPO operation)
Does it Works?
34BEAGLES… See the full description on the dataset page: https://huggingface.co/datasets/fblgit/simple-math.SimpleMath
🧮 SimpleMath 100K
SimpleMath 100K is a high-quality synthetic dataset of 100,000 basic arithmetic problems — no noise, no tricks, just clean and accurate math.
✅ Purpose
This was made for small AI models — not to struggle with complex math, but to get simple math right every time.
📦 Contents
75,000 numeric problems, evenly split:
18,750 addition (456 + 789 =)
18,750 subtraction (900 - 345 =)
18,750 multiplication (12 x 15 =)
18,750 division (144 / 12 =)… See the full description on the dataset page: https://huggingface.co/datasets/ProCreations/SimpleMath.Simple-MathSteps-90K
Introducing Simple-MathSteps-90K:
An open source dataset of 93,325 elementary math problems with step-by-step solutions and multiple choice answers. Designed to enhance mathematical reasoning in models ranging from 1B to 13B parameters.
Key Features
93,325 Math Problems: Generated by paraphrasing the AQuA-RAT dataset using Qwen3 4B Instruct 2507, with a focus on consistency and quality.
Detailed Step-by-Step Solutions: Clear reasoning that breaks down problems… See the full description on the dataset page: https://huggingface.co/datasets/Raymond-dev-546730/Simple-MathSteps-90K.simplemath-cot
🧮 SimpleMath-100k CoT
A chain-of-thought (CoT) extension of the
ProCreations/SimpleMath
dataset. Every one of the 100 000 algebra / arithmetic problems is paired with a
short, numbered reasoning trace (Step 1: … Step 2: …) that walks a language
model from the problem statement to the known-correct answer.
The traces in the Jupyter notebook are generated by
Qwen3.8-27B and then post-processed to strip formatting noise,
enforce sequential step numbering, and cap output at 1 000… See the full description on the dataset page: https://huggingface.co/datasets/alexfromapex/simplemath-cot.simple-math-DPO
Simple Math: 2+2=4 -1=3 (LoLo: Learning Only Logical Operations) DPO Pairs
Just like my teacher gave me homework, i thought maybe we can also add some of these basics on the trainings of our models.
It was created with very simple code that is in the repo, if you add more complex operations and so.. please share the code :D thank you
Current Code Version: 20240127.fblgit (A modification over @win10 for progressive and DPO operation)
Versions
27.01.24 First DPO… See the full description on the dataset page: https://huggingface.co/datasets/fblgit/simple-math-DPO.math-simple
🧮 Simple Math DB
This is a simple math dataset containing millions of basic arithmetic expressions.Each file represents a different operation — great for pretraining or arithmetic reasoning tasks.
⚠️ Training Notes
Models may overfit easily, since the pattern is always x (operator) y = (answer).It’s strongly recommended to train/test split the data rather than train on the whole thing.
📊 Dataset Info
Total lines: ~96,000,000
6 files (Addition… See the full description on the dataset page: https://huggingface.co/datasets/Roman190928/math-simple.simplemath-ita-sparse
Dataset Card for "simplemath-ita-sparse"
More Information needed
simple-math-qwen-7bSimple tasks from ProCreations/SimpleMath solved with Qwen/Qwen2.5-Math-7B-Instruct
simple_math_2_numbers_10mDAPO-Math-17k_simple_jjesimple-math-steps-7MSimple math expression solving with 3-6 operands and +-*/%^ operators, small powers and numbers between 1,1000 as operands.
A lot of the entries are incorrect, as it doesnt follow the BODMAS rule. I discovered the 10GB GLM dataset after this so this is abandoned.
The initial idea was to do a GLM type experiment.
Evaluate a model on word math problems
Train on equations
Check word math problem performance
The idea is to check if training on numbers only improves number understanding /… See the full description on the dataset page: https://huggingface.co/datasets/shb777/simple-math-steps-7M.simple_math-1MCreated using https://github.com/Ttimofeyka/numdatagen.
simple-mathsimple-math-x8_000_000simple_math_benchmark_multilingualsimplemath-400kfblgit_simple-math-DPO-PreferenceShareGPTUnroll-Qwen2.5-7B-Instruct_1754915847_eval_6a28_math500_simple-avg_num_prune_ffn_5_run-002
chengfu0118/Unroll-Qwen2.5-7B-Instruct_1754915847_eval_6a28_math500_simple-avg_num_prune_ffn_5_run-002
Precomputed model outputs for evaluation.
Evaluation Results
MATH500
Accuracy: 30.00%
Accuracy
Questions Solved
Total Questions
30.00%
150
500
simple-math-customrouter_PEFT_data_Math_simple_prompt_Qwen3-8Bu1-simple-math-2502Simple-Mathsimplemath-ita-sequential
Dataset Card for "simplemath-ita-sequential"
More Information needed
router_PEFT_data_Math_simple_prompt_Qwen3-1.7BUnroll-Qwen2.5-7B-Instruct_1754915889_eval_6a28_math500_simple-avg_num_prune_ffn_6_run-002
chengfu0118/Unroll-Qwen2.5-7B-Instruct_1754915889_eval_6a28_math500_simple-avg_num_prune_ffn_6_run-002
Precomputed model outputs for evaluation.
Evaluation Results
MATH500
Accuracy: 4.00%
Accuracy
Questions Solved
Total Questions
4.00%
20
500
Unroll-Qwen2.5-7B-Instruct_1754915759_eval_6a28_math500_simple-avg_num_prune_ffn_3_run-002
chengfu0118/Unroll-Qwen2.5-7B-Instruct_1754915759_eval_6a28_math500_simple-avg_num_prune_ffn_3_run-002
Precomputed model outputs for evaluation.
Evaluation Results
MATH500
Accuracy: 42.20%
Accuracy
Questions Solved
Total Questions
42.20%
211
500
Simple_Math_Logicv1Unroll-Qwen2.5-7B-Instruct_1754915716_eval_6a28_math500_simple-avg_num_prune_ffn_2_run-002
chengfu0118/Unroll-Qwen2.5-7B-Instruct_1754915716_eval_6a28_math500_simple-avg_num_prune_ffn_2_run-002
Precomputed model outputs for evaluation.
Evaluation Results
MATH500
Accuracy: 46.80%
Accuracy
Questions Solved
Total Questions
46.80%
234
500
Unroll-Qwen2.5-7B-Instruct_1754915801_eval_6a28_math500_simple-avg_num_prune_ffn_4_run-002
chengfu0118/Unroll-Qwen2.5-7B-Instruct_1754915801_eval_6a28_math500_simple-avg_num_prune_ffn_4_run-002
Precomputed model outputs for evaluation.
Evaluation Results
MATH500
Accuracy: 38.60%
Accuracy
Questions Solved
Total Questions
38.60%
193
500
router_PEFT_data_Math_simple_prompt_Qwen3-4B