datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
SYNTH-Swallow-Math-Code-Mix
Mixed dataset: SYNTH + SwallowMath-v2 + SwallowCode-v2
This high-signal, all synthetic dataset is a complete shuffled mix of the following four sources:
SYNTH ~63.5%
SwallowCode-v2 ~15.5%
SwallowMath-v2-textbook ~10.5%
SwallowMath-v2-qa ~10.0%
The motivation to provide this on HF was the need for a convenient, pre-shuffled merge of the highest quality synthetic / augmented datasets for small language model pre-training experiments as of… See the full description on the dataset page: https://huggingface.co/datasets/TMoC/SYNTH-Swallow-Math-Code-Mix.hero_run_4_math_codetest2_math_code_addedverified-math-code-17k
Verified Math & Code, 17,000 rows
A math and code instruction dataset where every single row was mechanically checked before it was
allowed in. Not filtered by a heuristic, not scored by a model. Checked.
Two layers of verification, one per domain:
Every math answer was compared against an independent gold answer by exact, numeric and
symbolic (SymPy) comparison. If the worked solution did not arrive at the gold answer, the row
was dropped.… See the full description on the dataset page: https://huggingface.co/datasets/manifesta/verified-math-code-17k.32b_exploit_seed_math_code_dedup_decontaminatemopd-math-code-mix
MOPD math+code mix
`train/`: math:code ≈ 1:1 平衡集（`math.parquet` + `code_*.parquet` shards）
`val/mopd_val_mix.parquet`: AIME24 全量 + MATH-500 子集 + Eurus code_validation 子集
路由字段：`ability ∈ {math, code}`
linear-programmingpreference_dataset_mixture2_and_safe_pku30k_and_argilla_math_and_ultra_code_for_preference_modelaptchat-v2-math-code-general-50kmath-ai-bench-sources-code
math-ai-bench-sources-code
A code benchmark evaluation dataset with 83,072 solution trajectories generated by state-of-the-art thinking models on coding benchmark problems.
Overview
Each entry is a long-form solution trajectory (chain-of-thought + final code) produced by a reasoning model on a held-out coding benchmark. Every trajectory carries a verified correct label, and every problem carries a correct_ratio (pass rate over all trajectories for that problem).… See the full description on the dataset page: https://huggingface.co/datasets/haowu89/math-ai-bench-sources-code.mopd-math-code-full-valmath_codemath_dataset_tinyamthinking_code_math_mixed_100kOpenThoughts3-full-filtered-math-no-codefiltered_sky_code_8k_math_10k_rubric_evidence_classifymath-code-adaptive-50kmath_reasoning_automated_problem_solving_with_code_track_3aptchat-v2-code-math-1kaptchat-v2-code-math-100kvietnamese-mathcoder
Dataset description
This dataset contains both English and Vietnamese samples.
Each sample has 3 solutions including a coding solution, an explanation solution and a direct solution. They are distinguished by the tokens <sol>, <exp> and <dir>, respectively.
For coding solution, execute the Python code wrapped inside ``` for the output result. DON'T TRUST the choice generated by model after coding, just execute the code and map it to the choice.
With this dataset & QLoRA on… See the full description on the dataset page: https://huggingface.co/datasets/thanhdath/vietnamese-mathcoder.math_dataset_smallaptchat-code-math-0.5kOHP-15k-mathcodemath-code-africa
Math & Code — African Context Dataset
Instruction-tuning dataset covering math and coding in African contexts: word problems with African currencies (UGX, KES, NGN), names, geography, and real-world scenarios (mobile money, market trading, farming); coding challenges for USSD systems, mobile money APIs, SMS gateways, agricultural data pipelines, and multilingual NLP — grounded via web search, generated with gemini-2.5-flash.
Dataset Details
Rows: 331
Regions… See the full description on the dataset page: https://huggingface.co/datasets/gimmy256/math-code-africa.math_datasetSFT-Math-Code
SFT-Math-Code
Generation
This dataset was created by running
uv run sft-math-code.py -H -U mikasenghaas
# sft-math-code.py
# /// script
# requires-python = ">=3.12"
# dependencies = ["datasets", "jinja2"]
# ///
import argparse
import json
import sys
from pathlib import Path
from typing import cast
from huggingface_hub import DatasetCard, DatasetCardData, whoami
fromdatasets import Dataset, DatasetDict, IterableDataset, load_dataset
def format_num(num: float | int… See the full description on the dataset page: https://huggingface.co/datasets/mikasenghaas/SFT-Math-Code.mt_bench_deepl_translated_fi_no_code_mathmath-tasks-datasetfiltered_sky_code_8k_math_10k_rubric_reasoning