datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
Magicoder-OSS-Instruct-75KThis is the OSS-Instruct dataset generated by gpt-3.5-turbo-1106 developed by OpenAI. Please pay attention to OpenAI's usage policy when adopting this dataset: https://openai.com/policies/usage-policies.
Magicoder-Evol-Instruct-110KA decontaminated version of evol-codealpaca-v1. Decontamination is done in the same way as StarCoder (bigcode decontamination process).
handy-dictation-editing
Handy dictation-editing corpus
Turns a raw dictated transcript into the text the speaker meant to write.
in : um so the meeting is uh moved to friday no wait thursday at three
out: The meeting is Thursday at three.
Three jobs at once, because they are not separable in speech: drop filler words,
repair punctuation and capitalisation, and — the hard one — when the speaker
changes their mind mid-sentence, delete the wording they abandoned and keep only
what they settled on.
Built… See the full description on the dataset page: https://huggingface.co/datasets/MagicNoThief/handy-dictation-editing.oxe-magic-soup-plus-nl-commands
OXE Magic Soup++ Natural Language Commands
Per-step natural-language instructions extracted from the datasets in the OpenVLA oxe_magic_soup_plus mixture of Open X-Embodiment (OXE). Source datasets were downloaded and preprocessed with rlds_dataset_mod, following OpenVLA, then processed with oxe-stats.
Files
One CSV per source dataset:
oxe-<dataset>_nl_only.csv — train split
oxe-<dataset>_validation_nl_only.csv — all non-train splits, merged
Columns… See the full description on the dataset page: https://huggingface.co/datasets/agnesluhtaru/oxe-magic-soup-plus-nl-commands.duplex-qa-refusal
duplex-qa-refusal
No dialogue in this set has been validated by a human.
Text-side augmentation of the moshika spoken-QA corpus so a full-duplex speech model can be trained to refuse a query when a mid-conversation text instruction tells it to, voice the reason the instruction gives, and then carry on normally. Two classes: policy (an existing benign query is declined for a stated reason; comes with an untouched accept twin sharing pair_id) and attack (a new user turn pivots to… See the full description on the dataset page: https://huggingface.co/datasets/MagicLuke/duplex-qa-refusal.Magicoder-Evol-Instruct-110K-sftThis dataset is a fork of https://huggingface.co/datasets/ise-uiuc/Magicoder-Evol-Instruct-110K.
It is just a version with the samples of CodeUltraFeedback filtered out.
Magicoder-OSS-Instruct-Rust-cleaned-3.9K
🦀 Magicoder-OSS-Instruct-Rust (3.9K Cleaned)
Magicoder-OSS-Instruct-Rust is a high-quality, syntax-verified dataset of 3,909 Rust coding instructions derived from real-world open-source GitHub projects.
This dataset is extracted from ise-uiuc/Magicoder-OSS-Instruct-75K, filtered specifically for Rust, and validated via in-memory compiler checks. No language translation was applied; the dataset remains in its original English format.
⚙️ Filtering and Verification… See the full description on the dataset page: https://huggingface.co/datasets/WrittenWithRust/Magicoder-OSS-Instruct-Rust-cleaned-3.9K.Magicoder-OSS-Instruct-Rust-TR-3.9K
🦀 Magicoder-OSS-Instruct-Rust-Turkish (3.9K)
Magicoder-OSS-Instruct-Rust-Turkish, WrittenWithRust/Magicoder-OSS-Instruct-Rust-cleaned-3.9K veri setindeki 3.909 adet sentaksı doğrulanmış İngilizce Rust instruction örneğinin tamamen Türkçe diline çevrilmesiyle oluşturulmuş yüksek kaliteli bir kod veri setidir.
Bu veri seti, Büyük Dil Modellerine (LLM) Türkçe Rust kodlama becerisi, problem çözme yeteneği ve karmaşık mimarileri açıklama kabiliyeti kazandırmak üzere Instruction… See the full description on the dataset page: https://huggingface.co/datasets/WrittenWithRust/Magicoder-OSS-Instruct-Rust-TR-3.9K.magicoder-cot-92k
Magicoder COT 92K
Chain-of-Thought distilled dataset for code generation tasks.
Dataset Description
This dataset contains 92,334 examples with added <think> reasoning traces, distilled from the original Magicoder-Evol-Instruct-110K dataset.
Source
Base dataset: ise-uiuc/Magicoder-Evol-Instruct-110K
Distillation model: Mistral-7B-Instruct-v0.2-AWQ
Method: vLLM inference with 50 parallel workers
Format
Each example contains:
instruction: The coding… See the full description on the dataset page: https://huggingface.co/datasets/domofon/magicoder-cot-92k.MagicBench
MagicBench: A Deception-Sensitive Cognitive Benchmark for LLMs
Project page | Code | Dataset
MagicBench is a deception-sensitive cognitive benchmark for language models built around magic-trick understanding. Rather than testing recall alone, it probes whether a model can reason about hidden causes, audience beliefs, violated expectations, uncertainty, and counterfactual changes in settings where the visible events are intentionally misleading.
Dataset Summary… See the full description on the dataset page: https://huggingface.co/datasets/hsiung/MagicBench.turkish-multi-turn-dialog-dataset
🇹🇷 Turkish Multi Turn Dialog Dataset — Türkçe Çok Turlu Sohbet Veri Seti
Türkçe doğal diyalog verisi — Tek turlu soru-cevap çiftlerinden, Gemini modeli kullanılarak üretilmiş çok turlu (multi-turn) konuşmalara dönüştürülmüş sentetik bir eğitim veri setidir.
📋 Veri Seti Özeti
Özellik
Değer
Dil
Türkçe
Konuşma başına tur sayısı
6 – 10 (rastgele)
Üretim modeli
gemini-3-flash-preview
Kaynak veri
LIMA (çevrilmiş veri setleri)
Amaç
SFT / Instruction Tuning… See the full description on the dataset page: https://huggingface.co/datasets/magibu/turkish-multi-turn-dialog-dataset.nullxes-train-vFINAL
nullxes-train-vFINAL
ShareGPT JSONL for NULLXES MGE CHIMERA vFINAL SFT (813 rows).
radon-test-code_generation
radon-test-code_generation
Description
Code generation test dataset for RADON model evaluation with programming prompts
Usage
Load Dataset
from datasets import load_dataset
dataset = load_dataset("MagistrTheOne/radon-test-code_generation")
print(dataset)
Use with RADON Model
from transformers import AutoModelForCausalLM, AutoTokenizer
# Load RADON model
model = AutoModelForCausalLM.from_pretrained("MagistrTheOne/RadonSAI")
tokenizer =… See the full description on the dataset page: https://huggingface.co/datasets/MagistrTheOne/radon-test-code_generation.radon-examples
radon-examples
Description
Usage examples and expected responses for RADON model with Russian and English samples
Usage
Load Dataset
from datasets import load_dataset
dataset = load_dataset("MagistrTheOne/radon-examples")
print(dataset)
Use with RADON Model
from transformers import AutoModelForCausalLM, AutoTokenizer
# Load RADON model
model = AutoModelForCausalLM.from_pretrained("MagistrTheOne/RadonSAI")
tokenizer =… See the full description on the dataset page: https://huggingface.co/datasets/MagistrTheOne/radon-examples.Magicoder-OSS-Instruct-75KThis is the OSS-Instruct dataset generated by gpt-3.5-turbo-1106 developed by OpenAI. Please pay attention to OpenAI's usage policy when adopting this dataset: https://openai.com/policies/usage-policies.
databricks-dolly-1k
Databricks Dolly 1k
1092 instruction examples taken from the original databricks/databricks-dolly-15k.
Filtered to open/closed/general QA category
Ready to plug straight into SFTTrainer, Unsloth, Llama-factory etc
Example
### Instruction:
When did Virgin Australia start operating?
### Context:
Virgin Australia, the trading name of Virgin Australia Airlines Pty Ltd ...
### Response:
Virgin Australia commenced services on 31 August 2000 as Virgin Blue, with two aircraft… See the full description on the dataset page: https://huggingface.co/datasets/MagicaNeko/databricks-dolly-1k.Banka_Diyalog_Magibu
Bankacılık/Finans Diyalog Veri Seti
Türkçe bankacılık ve finans müşteri asistanı için hazırlanmış diyalog veri seti. Fine-tuning amacıyla kullanılmak üzere oluşturuldu.
Veri Seti Yapısı
Veri seti train ve test olarak ikiye ayrılmıştır. Her satır şu alanları içerir:
system: sabit sistem talimatı
source: verinin kaynağı (handwritten ya da synthetic-gpt-4o-mini)
conversations: [{"role": "user"/"assistant", "content": "..."}] formatında diyalog
original_question:… See the full description on the dataset page: https://huggingface.co/datasets/Resad2173/Banka_Diyalog_Magibu.radon-test-long_context
radon-test-long_context
Description
Long context test dataset for RADON model evaluation with extended text samples
Usage
Load Dataset
from datasets import load_dataset
dataset = load_dataset("MagistrTheOne/radon-test-long_context")
print(dataset)
Use with RADON Model
from transformers import AutoModelForCausalLM, AutoTokenizer
# Load RADON model
model = AutoModelForCausalLM.from_pretrained("MagistrTheOne/RadonSAI")
tokenizer =… See the full description on the dataset page: https://huggingface.co/datasets/MagistrTheOne/radon-test-long_context.wikipedia-40-langs
Wikipedia 40 Languages
A curated multilingual dataset of Wikipedia articles spanning 40 languages with 812,000 articles total. Designed for multilingual NLP research, language modeling, and cross-lingual transfer learning.
Dataset Summary
This dataset contains Wikipedia articles from 40 languages, sampled and split into train/validation/test sets with a consistent 10:3:1 ratio per language. English and Turkish are overrepresented (10x more samples) to support focused… See the full description on the dataset page: https://huggingface.co/datasets/magibu/wikipedia-40-langs.radon-test-multilingual
radon-test-multilingual
Description
Multilingual test dataset for RADON model evaluation with Russian and English prompts
Usage
Load Dataset
from datasets import load_dataset
dataset = load_dataset("MagistrTheOne/radon-test-multilingual")
print(dataset)
Use with RADON Model
from transformers import AutoModelForCausalLM, AutoTokenizer
# Load RADON model
model = AutoModelForCausalLM.from_pretrained("MagistrTheOne/RadonSAI")
tokenizer =… See the full description on the dataset page: https://huggingface.co/datasets/MagistrTheOne/radon-test-multilingual.foodista
Foodista
Description
Foodista is a community-maintained site with recipes, food-related news, and nutrition information.
All content is licensed under CC BY.
Plain text is extracted from the HTML using a custom pipeline that includes extracting title and author information to include at the beginning of the text.
Additionally, comments on the page are appended to the article after we filter automatically generated comments.
Dataset Statistics… See the full description on the dataset page: https://huggingface.co/datasets/Magistr-Shuba/foodista.identity_finetune_magibu_q3
Identity Finetune Magibu Q3 Dataset
This dataset is designed for finetuning language models (specifically the Magibu series) to establish and maintain a consistent identity across Turkish and English languages.
Dataset Structure
The dataset is organized into two primary subsets using DatasetDict:
turkish: Contains training examples in Turkish focused on model identity, origins, and capabilities.
english: Contains corresponding training examples in English.… See the full description on the dataset page: https://huggingface.co/datasets/alibayram/identity_finetune_magibu_q3.magi-py
magi-py
Python-only subset of ise-uiuc/Magicoder-OSS-Instruct-75K.
Construction
from datasets import load_dataset
ds = load_dataset("ise-uiuc/Magicoder-OSS-Instruct-75K", split="train")
ds_py = ds.filter(lambda r: r["lang"] == "python")
# kept fields: {problem, solution}
Stats
38,284 examples (of 75K in the source)
Fields: problem (str), solution (str — typically a fenced ```python block)
License
MIT, inherited from the source dataset.