datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
lca-bug-localization
🏟️ Long Code Arena (Bug localization)
This is the benchmark for the Bug localization task as part of the
🏟️ Long Code Arena benchmark.
The bug localization problem can be formulated as follows: given an issue with a bug description and a repository snapshot in a state where the bug is reproducible, identify the files within the repository that need to be modified to address the reported bug.
The dataset provides all the required components for evaluation of bug localization… See the full description on the dataset page: https://huggingface.co/datasets/JetBrains-Research/lca-bug-localization.bugbounty-hunter-v1
Bug Bounty Hunter v1 - Fine-Tuning Dataset
The largest public dataset for training AI models on bug bounty hunting expertise.
Overview
52,743 instruction-response pairs covering 40+ vulnerability types across 6 categories, extracted from real-world bug bounty writeups and security research.
Dataset Summary
This dataset is designed to fine-tune language models to become expert bug bounty hunters. Each example contains a technical instruction and a… See the full description on the dataset page: https://huggingface.co/datasets/0xn0cta/bugbounty-hunter-v1.SWE-universe-repaired-bug-pilot-trajectories
SWE-universe repaired BugPilot trajectories
Combined trajectory artifacts for the Qwen3.6 + mini-swe-agent evaluation of VmaxRL/SWEUniverse-Repaired-Bugpilot.
This dataset contains one row per evaluated task in metadata.jsonl, plus per-task files under trajectories//. The combined set uses the main full eval and replaces the two original infra-failure rows with the clean infra rerun trajectories.
Summary:
rows: 804
effective attempts: 804
passes: 629
pass rate: 0.782338
infra… See the full description on the dataset page: https://huggingface.co/datasets/VmaxRL/SWE-universe-repaired-bug-pilot-trajectories.bugpilot-bugintro-lm-modify-gpt55-1k-oracleclean-443-20260506
BugPilot LM-Modify GPT-5.5 1k Oracle-Clean 443
This dataset contains the 443 task directories from the repaired LM-modify workspace that currently pass the oracle audit.
Source workspace: /data/augustine/demiurge/projects/experimental/training_swe_skrl_tinker/audits/lm_modify_target600_repair_workspace_20260506_v4
Source audit: full_oracle_postswaps_20260506_multinode32_c4
Export date: 2026-05-06
The directory layout matches the original task dataset layout: one task directory per… See the full description on the dataset page: https://huggingface.co/datasets/VmaxRL/bugpilot-bugintro-lm-modify-gpt55-1k-oracleclean-443-20260506.Bilge-Turkish-CoT-50K
Bilge: Turkish Chain-of-Thought Dataset (50K)
50,000 örneklik Türkçe Chain-of-Thought (CoT) reasoning fine-tuning veri seti.
Bilge, Türkçe büyük dil modellerinin adım adım düşünme (reasoning) kapasitesini
geliştirmek amacıyla hazırlanmış bir Chain-of-Thought veri setidir.
Veri setindeki her örnek, modelin önce <think> blokları içinde görünür bir
muhakeme süreci yürütmesini, ardından kullanıcıya yapılandırılmış ve detaylı
bir cevap vermesini öğretmek üzere tasarlanmıştır.
Bu… See the full description on the dataset page: https://huggingface.co/datasets/bugrabilge/Bilge-Turkish-CoT-50K.ierd-codeforces-subtle-bugs
IERD Codeforces subtle bugs
This public dataset contains 682 generated buggy C++ solutions for 682 Codeforces
problems. Each solution passes most tests in the frozen source corpus and fails from
one to five stored human or Hugging Face tests. The package also contains the frozen
manifest, provenance files, and aggregate reports from the final test generation
study.
Source and version
The problems, tests, and reference solution candidates come from… See the full description on the dataset page: https://huggingface.co/datasets/shivank21/ierd-codeforces-subtle-bugs.BuggedPythonLeetCode
Dataset Description
edit: fixed some bugs with datasets not handling all pyarrow types.
Dataset Summary
This dataset consists of Python coding problems from LeetCode, which have been bugged using the OpenBugger package. This dataset provides a unique opportunity to study the debugging process in a controlled and replicable environment.
For each correct code snippet, 15 bugged versions were attempted. For each succesfully bugged version, a corresponding question mimicking… See the full description on the dataset page: https://huggingface.co/datasets/NeuroDragon/BuggedPythonLeetCode.benchname-bug-localization
🥷 BenchName (Bug localization)
This is the benchmark for the Bug localization task as part of the
🥷 BenchName benchmark.
The bug localization problem can be formulated as follows: given an issue with a bug description and a repository snapshot in a state where the bug is reproducible, identify the files within the repository that need to be modified to address the reported bug.
The dataset provides all the required components for evaluation of bug localization approaches in… See the full description on the dataset page: https://huggingface.co/datasets/anon-iclr-submission/benchname-bug-localization.linux-kernel-bugfixes-diffs
🐧 Linux Kernel Bugfixes & Patches Dataset (Instruction-Tuned)
📖 Dataset Description
This dataset is a highly curated, instruction-tuned collection of problem-solution pairs extracted directly from the official Linux Kernel Git repository (torvalds/linux). It is specifically designed to train Large Language Models (LLMs) on low-level C programming, kernel architecture, memory management, and security vulnerability patching.
Unlike raw commit histories, this… See the full description on the dataset page: https://huggingface.co/datasets/switlydev/linux-kernel-bugfixes-diffs.bugbounty-hunter-v1
Bug Bounty Hunter v1 - Fine-Tuning Dataset
The largest public dataset for training AI models on bug bounty hunting expertise.
Overview
52,743 instruction-response pairs covering 40+ vulnerability types across 6 categories, extracted from real-world bug bounty writeups and security research.
Dataset Summary
This dataset is designed to fine-tune language models to become expert bug bounty hunters. Each example contains a technical instruction and a… See the full description on the dataset page: https://huggingface.co/datasets/ArkhAngelLifeJiggy/bugbounty-hunter-v1.sql-create-context-instruction
Overview
This dataset is built upon SQL Create Context, which in turn was constructed using data from WikiSQL and Spider.
There are 78,577 examples of natural language queries, SQL CREATE TABLE statements, and SQL Query answering the question using the CREATE statement as context. This dataset was built with text-to-SQL LLMs in mind, intending to prevent hallucination of column and table names often seen when trained on text-to-SQL datasets. The CREATE TABLE statement can often be… See the full description on the dataset page: https://huggingface.co/datasets/bugdaryan/sql-create-context-instruction.proximal-bug-injection-sft
Proximal Bug-Injection SFT
Supervised fine-tuning data for teaching a model to inject realistic, hard-to-find bugs into otherwise-correct code. The goal is to build coding RL environments where the introduced bugs resemble real human mistakes rather than trivial off-by-one or sign-flip errors.
The data is mined from real GitHub bug fixes: for each fix we take the after (correct) file as input and train the model to produce the before (buggy) version — i.e. the inverse of a… See the full description on the dataset page: https://huggingface.co/datasets/nreHieW/proximal-bug-injection-sft.BuggyVerilogFor usage, please take a look at https://github.com/CatIIIIIIII/VeriDebug.
cuda-to-rocm-wavefront-bugs
CUDA → ROCm Wavefront Bug Dataset
170 expert-curated examples of GPU kernel bugs that survive mechanical hipify translation and only manifest on AMD MI300X hardware (gfx942, wavefront-64).
Built for the ROCmPort AI project — a multi-agent pipeline that ports and optimizes CUDA kernels for AMD GPUs.
Why This Dataset Exists
hipify-perl and hipify-clang do a great job of mechanical API renaming (CUDA → HIP). But they cannot detect semantic bugs caused by AMD's larger… See the full description on the dataset page: https://huggingface.co/datasets/tazwarrrr/cuda-to-rocm-wavefront-bugs.prg-bench-buggy
PRG-Bench — buggy competitive programming solutions
1000 competitive programming problems, each paired with a correct C++ solution and a
variant containing exactly one injected defect. Every problem appears once — statements
are deduplicated by text, not by contest id, because Codeforces mirrors its harder
problems across the Div1 and Div2 runnings of a contest under two different ids.
Built to study whether reasoning models detect subtle defects in code, and whether that… See the full description on the dataset page: https://huggingface.co/datasets/cooperlee123/prg-bench-buggy.Bilge-Turkish-CoT-50K-Part2
Bilge: Turkish Chain-of-Thought Dataset (50K) — Part 2
İkinci 50.000 örneklik Türkçe Chain-of-Thought (CoT) reasoning fine-tuning veri seti.
Bilge Part 2, bugrabilge/Bilge-Turkish-CoT-50K
veri setinin devamı niteliğindeki ikinci 50.000 örneklik alt kümedir. İlk
bölümle örtüşmeyecek şekilde seçilmiş yeni örnekler içerir; iki bölüm
birlikte aynı 249K filtrelenmiş kümeden alınan toplam 100.000 örneklik bir
koleksiyon oluşturur.
İki bölüm birleştirildiğinde elde edilen 100.000… See the full description on the dataset page: https://huggingface.co/datasets/bugrabilge/Bilge-Turkish-CoT-50K-Part2.buggy-conversation-redo
Buggy Conversation Redo Dataset
This dataset contains conversations about code generation with potential security vulnerabilities.
It includes responses from different models with two variants:
original: Standard model responses
act_as_a_security_researcher: Responses where the model was prompted to act as a security researcher
Dataset Structure
Each record contains:
model_name: The name of the model used (e.g., gpt_oss_120b, command_a, etc.)
variant: Either "original"… See the full description on the dataset page: https://huggingface.co/datasets/regularpooria/buggy-conversation-redo.Hinglish-Everyday-Conversations-1M
Dataset Card for Hinglish Everyday Conversations Dataset
A synthetically created Hinglish-based dataset of 2 columns where every row represents a unique conversation between 2 people in Hinglish about Everyday Life Topics.
Use Model
Access the model made using this dataset: Tiny-Hinglish-Chat-21M
For more information about this model, its training process, or related resources, you can check the GitHub repository Tiny-Hinglish-Chat-21M-Scripts.
Dataset Details… See the full description on the dataset page: https://huggingface.co/datasets/buggiebug/Hinglish-Everyday-Conversations-1M.SyntheticCallStarters
SyntheticCallStarters
Starters to phone calls, ranging from fundraising requests to medical emergencies, generated by ChatGPT 4o mini.
BugWhisperer-dataset
Vulnerable Hardware Design Dataset
This repository contains a collection of vulnerable hardware design examples stored in a JSON Lines (JSONL) file format. Each line in the file represents a single vulnerability entry in JSON format, making it easy to process and analyze programmatically. Our top-performing model is now accessible on the Hugging Face platform for the research community. You can explore it here: [https://huggingface.co/shamstarek/Mistral-7B-instruct-Bug-Whisperer]… See the full description on the dataset page: https://huggingface.co/datasets/shamstarek/BugWhisperer-dataset.babylm-bug
BabyLM Dataset
Dataset Description
This dataset is part of the BabyLM multilingual collection.More information at: babylm.github.io/babybabellm
Dataset Summary
Language: bug
Script: Latn
Tier: 1M
Byte Premium Factor: 1.227860
Size (MB): 6.67
Expected Size (MB): 6.67
Number of Documents: 7,777
Total Tokens: 1,002,579
Tokenizer: separate by whitespace
Tokens Per Category
child-books: 41,174 tokens
padding: 961,405 tokens
Tokens Per… See the full description on the dataset page: https://huggingface.co/datasets/BabyLM-community/babylm-bug.smolified-bug2fix
🤏 smolified-bug2fix
Intelligence, Distilled.
This is a synthetic training corpus generated by the Smolify Foundry.
It was used to train the corresponding model crimsoncodie09/smolified-bug2fix.
📦 Asset Details
Origin: Smolify Foundry (Job ID: 4f8e9f7c)
Records: 643
Type: Synthetic Instruction Tuning Data
⚖️ License & Ownership
This dataset is a sovereign asset owned by crimsoncodie09.
Generated via Smolify.ai.
mirror-lca-bug-localization
🏟️ Long Code Arena (Bug localization)
This is the benchmark for the Bug localization task as part of the
🏟️ Long Code Arena benchmark.
The bug localization problem can be formulated as follows: given an issue with a bug description and a repository snapshot in a state where the bug is reproducible, identify the files within the repository that need to be modified to address the reported bug.
The dataset provides all the required components for evaluation of bug localization… See the full description on the dataset page: https://huggingface.co/datasets/alucent/mirror-lca-bug-localization.tasktrove-bugsinpy-v4-oracle17-apptainer-v1
TaskTrove BugsInPy v4: oracle-verified Apptainer subset v1
This derived Harbor release contains 17 of 479 upstream tasks. Every included reference solution is grounded in the official BugsInPy patch and selected by verifier execution. The remaining tasks are retained in the exclusion ledger; they are not silently discarded.
The release targets offline, rootless Apptainer on aarch64. Validation evidence is stored under validation/. Do not describe the full 479-task source as… See the full description on the dataset page: https://huggingface.co/datasets/laion/tasktrove-bugsinpy-v4-oracle17-apptainer-v1.