datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
CodeChat
CodeChat: Developer–LLM Conversations Dataset
Paper: https://arxiv.org/abs/2509.10402
GitHub: https://github.com/Software-Evolution-Analytics-Lab-SEAL/CodeChat
CodeChat is a large-scale dataset comprising 82,845 real-world developer–LLM conversations, containing 368,506 code snippets generated across more than 20 programming languages, derived from the WildChat (i.e., general Human-LLMs conversations dataset). The dataset enables empirical analysis of how developers… See the full description on the dataset page: https://huggingface.co/datasets/Suzhen/CodeChat.code_contests_instruct
Dataset Card for "code_contests_instruct"
The deepmind/code_contests dataset formatted as markdown-instruct for text generation training.
There are several different configs. Look at them. Comments:
flesch_reading_ease is computed on the description col via textstat
hq means that python2 (aka PYTHON in language column) is dropped, and keeps only rows with flesch_reading_ease 75 or greater
min-cols drops all cols except language and text
possible values for language are {'CPP'… See the full description on the dataset page: https://huggingface.co/datasets/BEE-spoke-data/code_contests_instruct.CodeChat-V2.0
CodeChat: Developer–LLM Conversations Dataset
Paper: https://arxiv.org/abs/2509.10402
GitHub: https://github.com/Software-Evolution-Analytics-Lab-SEAL/CodeChat
CodeChat V2.0 is a large-scale dataset comprising 587,568 real-world developer–LLM conversations, derived from the WildChat dataset.
The V2.0 statistics below match Table I of the CASCON 2026 camera-ready paper, Developer-LLM Conversations: An Empirical Study of Interactions and Generated Code Quality.… See the full description on the dataset page: https://huggingface.co/datasets/Suzhen/CodeChat-V2.0.agent-sft-stitch-zh-tts-taste-codec-chat-sample
Gemma 4 E2B Taste-S multi-turn codec SFT
This dataset contains 37,362 complete Traditional Chinese agent
dialogues selected from voidful/agent-sft-stitch-zh-tts. It covers
229,434 synthesized speech segments, approximately
520.5 hours of audio before codec extraction.
Every assistant speech segment is represented without Gemma native audio tags:
<SAY> text_token <a_code> <b_code> ... <p_code> ... </SAY>
The first assistant output starts immediately with <SAY>.
[SOPR]...[EOPR]… See the full description on the dataset page: https://huggingface.co/datasets/voidful/agent-sft-stitch-zh-tts-taste-codec-chat-sample.code-commande-publique
Code de la commande publique, non-instruct (2025-09-20)
The objective of this project is to provide researchers, professionals and law students with simplified, up-to-date access to all French legal texts, enriched with a wealth of data to facilitate their integration into Community and European projects.
Normally, the data is refreshed daily on all legal codes, and aims to simplify the production of training sets and labeling pipelines for the development of free, open-source… See the full description on the dataset page: https://huggingface.co/datasets/louisbrulenaudet/code-commande-publique.codecomplex
CodeComplex Dataset
Dataset Description
CodeComplex consists of 4,200 Java codes submitted to programming competitions by human programmers and their complexity labels annotated by a group of algorithm experts.
How to use it
You can load and iterate through the dataset with the following two lines of code:
from datasets import load_dataset
ds = load_dataset("codeparrot/codecomplex", split="train")
print(next(iter(ds)))
Data Structure… See the full description on the dataset page: https://huggingface.co/datasets/codeparrot/codecomplex.code_contest_instruct_cppcode_clippyThis dataset was generated by selecting GitHub repositories from a large collection of repositories. These repositories were collected from https://seart-ghs.si.usi.ch/ and Github portion of [The Pile](https://github.com/EleutherAI/github-downloader) (performed on July 7th, 2021). The goal of this dataset is to provide a training set for pretraining large language models on code data for helping software engineering researchers better understand their impacts on software related tasks such as autocompletion of code. The dataset is split into train, validation, and test splits. There is a version containing duplicates (209GBs compressed) and ones where exact duplicates (132GBs compressed) are removed. Contains mostly JavaScript and Python code, but other programming languages are included as well to various degrees.code_contest_processed
Dataset Card for Code Contest Processed
Dataset Summary
This dataset is created by processing code_contest dataset from Deepmind. It is a competitive programming dataset for machine-learning. Read more about dataset at original source.
Columns Description
id : unique string associated with a problem
description : problem description
code : one correct code for the problem
language : programming language used for code
test_samples : contains inputs and their… See the full description on the dataset page: https://huggingface.co/datasets/iamtarun/code_contest_processed.CodeCompass
CodeCompass: A Benchmark for Code Generation
Paper: Rethinking Verification for LLM Code Generation: From Generation to Testing
Description
CodeCompass is a rigorous benchmark designed to evaluate the code generation capabilities of Large Language Models (LLMs). It comprises a comprehensive collection of programming problems sourced from competitive platforms, offering a standardized framework for assessing algorithmic reasoning, problem-solving, and code synthesis in a… See the full description on the dataset page: https://huggingface.co/datasets/opencompass/CodeCompass.code-civil
Code civil, non-instruct (2025-09-20)
The objective of this project is to provide researchers, professionals and law students with simplified, up-to-date access to all French legal texts, enriched with a wealth of data to facilitate their integration into Community and European projects.
Normally, the data is refreshed daily on all legal codes, and aims to simplify the production of training sets and labeling pipelines for the development of free, open-source language models based… See the full description on the dataset page: https://huggingface.co/datasets/louisbrulenaudet/code-civil.code-collectivites-territoriales
Code général des collectivités territoriales, non-instruct (2025-09-20)
The objective of this project is to provide researchers, professionals and law students with simplified, up-to-date access to all French legal texts, enriched with a wealth of data to facilitate their integration into Community and European projects.
Normally, the data is refreshed daily on all legal codes, and aims to simplify the production of training sets and labeling pipelines for the development of free… See the full description on the dataset page: https://huggingface.co/datasets/louisbrulenaudet/code-collectivites-territoriales.code-consommation
Code de la consommation, non-instruct (2025-09-20)
The objective of this project is to provide researchers, professionals and law students with simplified, up-to-date access to all French legal texts, enriched with a wealth of data to facilitate their integration into Community and European projects.
Normally, the data is refreshed daily on all legal codes, and aims to simplify the production of training sets and labeling pipelines for the development of free, open-source language… See the full description on the dataset page: https://huggingface.co/datasets/louisbrulenaudet/code-consommation.code-construction-habitation
Code de la construction et de l'habitation, non-instruct (2025-09-20)
The objective of this project is to provide researchers, professionals and law students with simplified, up-to-date access to all French legal texts, enriched with a wealth of data to facilitate their integration into Community and European projects.
Normally, the data is refreshed daily on all legal codes, and aims to simplify the production of training sets and labeling pipelines for the development of free… See the full description on the dataset page: https://huggingface.co/datasets/louisbrulenaudet/code-construction-habitation.codecl
Context Learning for Software Engineering
Code
•
Paper
Dataset Details
Dataset Sources [optional]
Code Generation: LeetCode
Code Summary: Github
Code Review: Github
Patch Correctness Assessment: Defects4J
Uses
Evaulate LLMs on context learning ability under Software Engineering tasks.
Citation
@misc{hu2026cl4secontextlearningbenchmark,
title={CL4SE: A Context Learning Benchmark For Software Engineering Tasks}… See the full description on the dataset page: https://huggingface.co/datasets/tomhu/codecl.CodeChat-Instruct-v1
CodeChat-Instruct-v1
CodeChat-Instruct-v1 is a synthetic coding instruction dataset designed for supervised fine-tuning of language models on programming-related conversations. It includes diverse coding tasks such as code review, code improvement, complexity analysis, edge-case discussion, code explanation, library/API usage, refactoring guidance. The dataset is suitable for training coding assistants, educational programming tutors, and general-purpose code LLMs with strong… See the full description on the dataset page: https://huggingface.co/datasets/kd13/CodeChat-Instruct-v1.critic-sft-cwm-only
critic-sft-cwm-only
The CWM-only SFT corpus from Steer, Don't Solve: Training Small Critic Models for Large Code Agents. It trains Qwen3-8B-Critic-SFT-CWM-only, the CWM-only arm of the corpus ablation in Table 3.
Each record is one critique point: a CWM-32B trajectory up to some step, followed by the critique that Claude Opus 4.6 wrote for it under the paper's high-level prompt. The main corpus, critic-sft-cwm-qwen, is this set plus 1,915 records from Qwen3-Next trajectories.… See the full description on the dataset page: https://huggingface.co/datasets/code-critic-model/critic-sft-cwm-only.code_contest_python3_alpaca
Dataset Card for Code Contest Processed
Dataset Summary
This dataset contains coding contest questions and their solution written in Python3.
This dataset is created by processing code_contest dataset from Deepmind. It is a competitive programming dataset for machine-learning. Read more about dataset at original source.
Columns Description
id : unique string associated with a problem
description : problem description
code : one correct code for the problem… See the full description on the dataset page: https://huggingface.co/datasets/iamtarun/code_contest_python3_alpaca.Code-CoT
G2RPO-A Code-CoT
Code-CoT contains 1,086 Python programming problems paired with generated chain-of-thought (CoT) trajectories and executable test cases. It is the coding guidance dataset released with G²RPO-A: Guided Group Relative Policy Optimization with Adaptive Guidance.
Paper (ACL 2026) · Code · Source dataset · Math-Curriculum-1K
Dataset at a glance
Property
Value
Task
Python code generation and reasoning
Intended use
Guidance for G²RPO-A and… See the full description on the dataset page: https://huggingface.co/datasets/G2RPO-A/Code-CoT.critic-sft-cwm-only-detailed-prompt
critic-sft-cwm-only-detailed-prompt
The detailed-prompt SFT corpus from Steer, Don't Solve: Training Small Critic Models for Large Code Agents. It trains Qwen3-8B-Critic-SFT-Detailed-Prompt, the comparison arm of the prompt ablation in Table 4.
Each record is one critique point: a CWM-32B trajectory up to some step, followed by the critique that Claude Opus 4.6 wrote for it. The difference from critic-sft-cwm-only is the teacher prompt. Here the teacher used the detailed prompt… See the full description on the dataset page: https://huggingface.co/datasets/code-critic-model/critic-sft-cwm-only-detailed-prompt.critic-sft-cwm-qwen
critic-sft-cwm-qwen
The main SFT corpus from Steer, Don't Solve: Training Small Critic Models for Large Code Agents. It trains both Qwen3-8B-Critic-SFT and Qwen3-4B-Critic-SFT.
Each record is one critique point: a coding agent's trajectory up to some step, followed by the critique that Claude Opus 4.6 wrote for it. The teacher was prompted with the paper's high-level prompt, which asks for error detection and one or two sentences of guidance and forbids code and commands in the… See the full description on the dataset page: https://huggingface.co/datasets/code-critic-model/critic-sft-cwm-qwen.Nemotron-Personas-Korea
Nemotron-Personas-Korea
우리나라 실제 분포에 기반한 합성 페르소나를 위한 복합 AI 시스템
A compound AI approach to personas grounded in real-world distributions
데이터셋 개요 (Overview)
Nemotron-Personas-Korea는 대한민국의 실제 인구통계학적·지리적·성격 특성 분포를 기반으로 합성된 오픈소스 페르소나 데이터셋(CC BY 4.0)으로, 우리나라 인구의 다양성과 특성을 폭넓게 반영하도록 설계되었습니다. 이는 최초의 대규모 우리말 페르소나 데이터셋이며, 이름, 성별, 나이, 혼인 상태, 교육 수준, 직업, 거주 지역 등의 속성을 실제 대한민국 통계청(KOSIS), 대법원, 국민건강보험공단, 농촌경제연구원, NAVER Cloud 통계 자료를 기반으로 합성하였습니다.
Nemotron-Personas-Korea는… See the full description on the dataset page: https://huggingface.co/datasets/codecainecowboy/Nemotron-Personas-Korea.code-contract-repair
APIContractRepair
APIContractRepair is a provenance-tracked instruction-tuning dataset for software engineers and code-model researchers who need contract-faithful, minimal repairs with tests that distinguish a broken implementation from its fix. Magicoder-OSS-Instruct-75K supplies real function-identifier seeds, but it does not provide these documented contracts, deliberately buggy implementations, minimal corrected implementations, or paired regression tests. This release… See the full description on the dataset page: https://huggingface.co/datasets/skonml/code-contract-repair.CodeContests_OnlyPythonDataThis dataset is based on the "code_contests" dataset by DeepMind, licensed under CC BY 4.0.
Original dataset available at: https://huggingface.co/datasets/deepmind/code_contests
このデータセットは、deepmind/code_contestsからAtCoderのデータのみを抽出し、Pythonで正解として提出された最初のコードを取得して格納したものです。これにより、教師あり学習に適した形に整えられています。
This dataset is created by extracting only the data from AtCoder within deepmind/code_contests and retrieving the first Python code submission marked as correct. As a result, it has been formatted to be… See the full description on the dataset page: https://huggingface.co/datasets/SSK-DNB/CodeContests_OnlyPythonData.code-contests-noblock
code-contests-noblock
Harbor-format competitive-programming RL tasks (8,728 tasks) derived from CodeContests.
Each task is a Harbor task binary (gzip tar) stored in tasks.parquet with columns
path (str, <task_id>.tar.gz) and task_binary (binary). Each tar contains
instruction.md, task.toml, environment/Dockerfile, and a tests/ verifier
(test.sh, test_state.py, test_data.json).
v2 (current) — verifier reward-recording fix
Resolves a silent reward-distribution… See the full description on the dataset page: https://huggingface.co/datasets/DCAgent/code-contests-noblock.GLM-5.1-Reasoning-1M-Cleaned
GLM-5.1-Reasoning-1M-Cleaned
GLM-5.1-Reasoning-1M-Cleaned is a cleaned and reformatted derivative of Kassadin88/GLM-5.1-1000000x. It preserves the original four-subset layout (main, PHD-Science, Multilingual-STEM, Math) while converting every example into a unified SFT-ready schema with explicit conversations, input, output, domain, and meta fields.
This release was prepared from the original dataset published by Kassadin88.
Summary
Teacher model in the data: GLM-5.1… See the full description on the dataset page: https://huggingface.co/datasets/codecainecowboy/GLM-5.1-Reasoning-1M-Cleaned.codecomplex
CodeComplex Dataset
Dataset Description
CodeComplex consists of 4,200 Java codes submitted to programming competitions by human programmers and their complexity labels annotated by a group of algorithm experts.
How to use it
You can load and iterate through the dataset with the following two lines of code:
from datasets import load_dataset
ds = load_dataset("codeparrot/codecomplex", split="train")
print(next(iter(ds)))
Data Structure… See the full description on the dataset page: https://huggingface.co/datasets/HlebYakhnitski/codecomplex.codecontests-textbooks-dp-v1This dataset is a synthetic collection designed for algorithmic problem-solving, particularly in the dynamic programming domain. It is inspired by problems from the DeepMind/code_contests dataset, ensuring authenticity and relevance to competitive programming and algorithmic challenges.
The dataset includes detailed problem statements, input-output specifications, constraints, and illustrative test cases. Each example mirrors real-world scenarios, providing not only the problem but also… See the full description on the dataset page: https://huggingface.co/datasets/eo-belyakin/codecontests-textbooks-dp-v1.critic-sft-qwen-only
critic-sft-qwen-only
The Qwen-only SFT corpus from Steer, Don't Solve: Training Small Critic Models for Large Code Agents. It trains Qwen3-8B-Critic-SFT-Qwen-only and Qwen3-4B-Critic-SFT-Qwen-only, the Qwen-only arms of the corpus ablation in Table 3.
Each record is one critique point: a Qwen3-Next-80B-A3B-Instruct trajectory up to some step, followed by the critique that Claude Opus 4.6 wrote for it under the paper's high-level prompt. The main corpus, critic-sft-cwm-qwen, is… See the full description on the dataset page: https://huggingface.co/datasets/code-critic-model/critic-sft-qwen-only.code-communes
Code des communes, non-instruct (2025-09-20)
The objective of this project is to provide researchers, professionals and law students with simplified, up-to-date access to all French legal texts, enriched with a wealth of data to facilitate their integration into Community and European projects.
Normally, the data is refreshed daily on all legal codes, and aims to simplify the production of training sets and labeling pipelines for the development of free, open-source language models… See the full description on the dataset page: https://huggingface.co/datasets/louisbrulenaudet/code-communes.