datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
LeetCodeDataset
LeetCodeDataset
LeetCodeDataset is a dataset consists of Python leetcode problems that can be used for LLM training and evaluation.
💻 GitHub
📄 LeetCodeDataset: A Temporal Dataset for Robust Evaluation and Efficient Training of Code LLMs
📄 Policy Filtration for RLHF to Mitigate Noise in Reward Models
leetcodetw-leetcode
Dataset Card for tw-leetcode
A curated Traditional Chinese LeetCode solution dataset with high-efficiency answers (Beats 100%), structured explanation in "Top Concept → Step Implement → Complexity Analysis" style, updated daily.
Dataset Details
Dataset Description
tw-leetcode 是一個針對 LeetCode 題目的繁體中文資料集，內容包含高效能程式解法、完整的解題思路，以及時間與空間複雜度分析。每份題解都經由人工清洗與優化，並依循「Top Concept → Step Implement → Complexity Explanation」的結構撰寫，方便機器學習模型或人類讀者理解程式邏輯的推理過程。
本資料集適合作為：… See the full description on the dataset page: https://huggingface.co/datasets/twinkle-ai/tw-leetcode.grpo-qwen3-1.7b-nemotron-leetcode-clean-3.2k-bs32-n8-verl091-epoch2-146102-rollouts
Coding RL rollouts
grpo_Qwen3-1.7B_Nemotron-LeetCode-clean-3.2k_bs32_n8_seqs16_32k_epoch2_verl091
One verified gzip JSONL shard per training step; 256 responses per shard.
LCB binary grading after thinking, without an EOS gate.
leetcode
LeetCode multilingual benchmark dataset
LeetCode problems for the msl-multilingual-self-learning benchmark, flattened to
one row per (problem, language) in 9 languages, stored as data/<split>/<lang>-NNN.jsonl.
Each row has the interface for its language (from LeetCode's code snippets),
the shared canonical_tests (Python asserts from newfacade/LeetCodeDataset),
the problem_description (from the LeetCode page) and metadata.
Tests that break the problem's Constraints, do not fit a… See the full description on the dataset page: https://huggingface.co/datasets/neulab/leetcode.grpo-qwen3-1.7b-nemotron-leetcode-clean-3.2k-bs32-n8-verl091-146102-rollouts
Coding GRPO rollouts
grpo_Qwen3-1.7B_Nemotron-LeetCode-clean-3.2k_bs32_n8_seqs16_32k_1epoch_verl091
One verified gzip JSONL shard per training step; 256 responses per shard.
LCB binary grading after thinking, without an EOS gate.
tigerbot-kaggle-leetcodesolutions-en-2kTigerbot 基于leetcode-solutions数据集，加工生成的代码类sft数据集
原始来源：https://www.kaggle.com/datasets/erichartford/leetcode-solutions
Usage
import datasets
ds_sft = datasets.load_dataset('TigerResearch/tigerbot-kaggle-leetcodesolutions-en-2k')
leetcode-complete
Complete LeetCode Problems Dataset
This dataset contains a comprehensive collection of LeetCode problems (including premium) with AI-generated solutions in JSONL format. It is regularly updated to include new problems as they are added to LeetCode.
Splits
The dataset is divided into the following splits:
train: Contains approximately 80% of the problems for training
validation: Contains approximately 10% of the problems for validation
test: Contains approximately… See the full description on the dataset page: https://huggingface.co/datasets/whiskwhite/leetcode-complete.leetcode_code_generationleetcode1000leetcodeleetcode-python-dataset
leetcode-python-dataset
Code for building and publishing the justindal/leetcode-python-dataset dataset on Hugging Face.
Merges two open-source LeetCode datasets into a unified schema with consistent formatting, field normalisation, and solution validation.
Dataset
Split
Rows
Source
train
2856
newfacade + greengerong
valid
310
slug-group split from train
test
228
newfacade only
Schema
default config (training)
Each row is a… See the full description on the dataset page: https://huggingface.co/datasets/justindal/leetcode-python-dataset.LeetCode-Contest
LeetCode Contest Benchmark
A new benchmark for evaluating Code LLMs proposed by DeepSeek-Coder, which consists of the latest algorithm problems of different difficulties.
Usage
git clone https://github.com/deepseek-ai/DeepSeek-Coder.git
cd Evaluation/LeetCode
# Set the model or path here
MODEL="deepseek-ai/deepseek-coder-7b-instruct"
python vllm_inference.py --model_name_or_path $MODEL --saved_path output/20240121-Jul.deepseek-coder-7b-instruct.jsonl
python… See the full description on the dataset page: https://huggingface.co/datasets/TechxGenus/LeetCode-Contest.leetcode_problem_solutionThis dataset contains: problems and solutions in Leetcode, crawled from: https://github.com/AnasImloul/Leetcode-Solutions
The format of data:
title: title of the problem
algo_input: the description of the problem
solution_py: the solution in Python
solution_js: the solution in Js
solution_java: the solution in Java
solution_c: the solution in C
leetcode-python-distill
leetcode-python-distill
Distill version of the justindal/leetcode-python-dataset dataset.
Includes thinking traces from Deepseek 3.2
Sources
newfacade/LeetCodeDataset (Apache 2.0)
greengerong/leetcode (MIT)
Citation
newfacade/LeetCodeDataset:
@misc{xia2025leetcodedatasettemporaldatasetrobust,
title={LeetCodeDataset: A Temporal Dataset for Robust Evaluation and Efficient Training of Code LLMs},
author={Yunhui Xia and Wei Shen and Yan Wang and Jason… See the full description on the dataset page: https://huggingface.co/datasets/justindal/leetcode-python-distill.leetcode-rosettaleetcodeLeetcode_dataset_llmLeetCodeleetcode_datasetLeetCode-Contest-zh
LeetCode Contest Benchmark
A new benchmark for evaluating Code LLMs proposed by DeepSeek-Coder, which consists of the latest algorithm problems of different difficulties.
Usage
git clone https://github.com/deepseek-ai/DeepSeek-Coder.git
cd Evaluation/LeetCode
# Set the model or path here
MODEL="deepseek-ai/deepseek-coder-7b-instruct"
python vllm_inference.py --model_name_or_path $MODEL --saved_path output/20240121-Jul.deepseek-coder-7b-instruct.jsonl
python… See the full description on the dataset page: https://huggingface.co/datasets/TechxGenus/LeetCode-Contest-zh.Dans-Codemaxx-LeetCodeLeetCode_with_Solutionsleetcode-performance
Dataset card for Leetcode Performance Dataset
LeetCode_Style_QAcleaned-humaneval-leetcode
Cleaned HumanEval-LeetCode Dataset
Dataset Description
这是一个经过预处理的 LeetCode 代码生成数据集，源自 ShwStone/humaneval-leetcode。
Data Preprocessing (清洗逻辑)
为了直接用于大模型 SFT 训练，原始数据经过了以下清洗：
Type Hinting: 根据代码内容，自动注入了 typing (List, Dict, etc.) 和 collections (deque, defaultdict) 的引用。
Class Removal: 移除了 class Solution 外壳，将代码转换为纯函数形式。
Refactoring: 移除了 self 参数，修复了缩进。
Docstring Repair: 修复了 doctest 中的函数调用名称。
Data Structure
task_id: 题目唯一标识符。
prompt:… See the full description on the dataset page: https://huggingface.co/datasets/Zedekiah557/cleaned-humaneval-leetcode.leetcodeThis dataset contains python solutions for various Leetcode problems, scraped from different posts by users from the solutions tab.
LogicStack-LeetCodeextract from LogicStack-LeetCode
公众号「宫水三叶的刷题日记」刷穿 LeetCode 系列文章源码
包括 编程题目、解析、tag、题目url
根据 leetcode 原始题目网页，修正了一些 文件名 和 文件内容 中标注的难度不一致的文件样本
humaneval-leetcode
HumanEval-LeetCode Input-Only
本数据集由 LeetCode 题目自动转换而来，目标是生成 HumanEval 风格的 Input-Only 提示（prompt）。每条样本包含题目描述与函数签名，描述中的 Input/Output 被转换成 >>> func(...) / >>> output 的交互式示例，便于直接用于代码生成评测或训练。
数据格式
每行是一个 JSON 对象（JSONL）：
task_id：题目唯一标识
prompt：包含函数签名与 docstring 的完整提示
示例字段：
task_id: string
prompt: string
生成流程（概述）
读取 LeetCode 题目（题目描述 + starter code）。
提取函数名。
将题目描述中的 Input/Output 转为：
>>> func(inputs...)
>>> output
输出为 JSONL。
数据划分
train.jsonl
test.jsonl… See the full description on the dataset page: https://huggingface.co/datasets/ShwStone/humaneval-leetcode.LeetCode_Conversations139 Code Conversations generated from LeetCode Questions including official answers.
Totals to 968 total messages from USER and SYSTEM.
Includes:
Generation of the solutions
Conversions into other programming languages
Adjustments to the Code
Generating tests
Conversations generated with GPT4/GPT4-Turbo