datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
explicit-edit-benchmark
Explicit Edit Benchmark
226 deterministic exact-edit tasks, run by different agents, harnesses, models and configurations. Every observation records what the harness did and whether the resulting files matched byte for byte.
Source code and benchmark runner: GitHub — Explicit Edit Benchmark
Open the interactive Explorer to compare agents, harnesses, models, versions, reasoning modes, correctness, recovery, time, cost and tokens.
Leaderboard by model route
Score v2… See the full description on the dataset page: https://huggingface.co/datasets/alexshpunt/explicit-edit-benchmark.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.codeforces-editorial-full-2026-07-16
Codeforces Editorial Full - 2026-07-16
A deterministic Plan-CRL-compatible materialization of open-r1/codeforces, pinned to revision fbe3f6e903ee854eec2e69e9d96d0306cde59baf.
Size
train: 9556 problems
test: 468 problems
total: 10024 unique problems
safe fixed-output Plan-CRL evaluation rows: 1479
safe fixed-output rows with an editorial: 646
The upstream snapshot contains 10,024 unique problems, not 11,000+. The larger counts sometimes quoted for this corpus… See the full description on the dataset page: https://huggingface.co/datasets/1337xyz1337xyz/codeforces-editorial-full-2026-07-16.EditPackFT
EditPackFT
EditPackFT is a dataset built for training LLMs on the task of instructional code editing. The mail columns are:
old_contents the code before the edit
instruction the instruction to transform the before code into the after code
new_contents the code after the edit
content a pre-formatted training window that can be used to train an LLM with prompts in the format of: <before><instruction><after>
This dataset has been filtered from CommitPackFT. For more detail, see our… See the full description on the dataset page: https://huggingface.co/datasets/nuprl/EditPackFT.sgu-editorial
ACM SGU Competitive Programming Solutions with LLM Enhancement
This dataset contains solutions to ACM SGU (Saratov State University) competitive programming problems, enhanced with detailed editorials and reasoning explanations generated using advanced language models. The full page about the project is here.
Overview
The dataset consists of two main components:
Original Solutions: Competitive programming solutions to SGU problems in C++ or Python.
Enhanced… See the full description on the dataset page: https://huggingface.co/datasets/radoslav11/sgu-editorial.taskweft-fbd-editscore-train
taskweft-fbd-editscore-train
Intents and the IEC 61131-3 Function Block Diagrams that carry them out, as an
EditScore-shaped corpus: one root row per intent, three candidates per row (rank1 the
reference diagram, rank3 one that compiles and does the wrong thing, rank5 one the
compiler refuses), and one score row per candidate from the compiler and a runner that performed the plan. Every row is
constructed from a template and a seed, so the labels are true by construction and
the… See the full description on the dataset page: https://huggingface.co/datasets/chibifire/taskweft-fbd-editscore-train.Nano3D-Edit-100k
Nano3D-Edit-100k
This dataset is the official data release for Nano3D, a training-free framework for precise and coherent 3D object editing without masks.
Paper: Nano3D: A Training-Free Approach for Efficient 3D Editing Without MasksProject Page: https://jamesyjl.github.io/Nano3D/
Nano3D integrates FlowEdit into TRELLIS to perform localized 3D edits guided by front-view renderings, and introduces Voxel/Slat-Merge strategies to preserve structural consistency between edited and… See the full description on the dataset page: https://huggingface.co/datasets/yejunliang23/Nano3D-Edit-100k.EditEval
EditEval: The Instruction-Based Benchmark for Text Improvements
This dataset contains the EditEval benchmark data, converted to JSONL and organized by task/dataset.
Subsets
Config
Task
Examples
jfleg
Fluency
1,501
asset
Simplification
2,359
turk
Simplification
2,359
iterater
Mixed (all tasks)
621
iterater_fluency
Fluency
203
iterater_clarity
Clarity
342
iterater_coherence
Coherence
76
stsb_multi_mt
Paraphrasing
153
wnc
Neutralization
1,700… See the full description on the dataset page: https://huggingface.co/datasets/bzz2/EditEval.Medical_Intelligence_Dataset_76k_2026_Edition
🏥 Medical Intelligence Dataset · 76k · 2026 Edition
Production-ready medical AI dataset for training diagnosis, treatment reasoning, and doctor-patient conversational systems.
76,000 engineered (not collected) English Q&A pairs — covering 620+ diseases,
438+ FDA-approved drugs, and real patient-doctor conversations. Built with a
5-stage quality pipeline. Commercial-safe Apache 2.0.
Created by Huzefa Nalkheda Wala — AI Product
Engineer & Medical AI Researcher · Creator of the… See the full description on the dataset page: https://huggingface.co/datasets/huzaifa525/Medical_Intelligence_Dataset_76k_2026_Edition.Twin-2K-500_edit
Twin-2K-500 Dataset
This dataset Twin-2K-500 contains comprehensive persona information from a representative sample of 2,058 US participants, providing rich demographic and psychological data. The dataset is specifically designed for building digital twins for LLM simulations.
More information on how to use this dataset can be found in our Documentation and GitHub repository.
Details on how the dataset was generated are available in our Paper.
Dataset Creation… See the full description on the dataset page: https://huggingface.co/datasets/Shar999/Twin-2K-500_edit.codeforces-editorial-elo-512-2026-04-28
Codeforces Editorial ELO 512 - 2026-04-28
A 512-example subset sampled from open-r1/codeforces (verifiable, train) for Plan-CRL Codeforces feedback experiments that need a non-empty trusted reference field.
Important: open-r1/codeforces does not expose code reference solutions. This subset fills reference_solution from the source dataset's editorial field. Treat it as a natural-language editorial/rationale, not canonical reference code.
Selection seed: 20260428.
Filtering:
rating… See the full description on the dataset page: https://huggingface.co/datasets/1337xyz1337xyz/codeforces-editorial-elo-512-2026-04-28.edith-marvel-corpus
EDITH Marvel corpus
202,171 cleaned records - characters, teams, locations, items, events, comics -
derived from Marvel Database (marvel.fandom.com) and English Wikipedia, crawled
2026-09-04 via the MediaWiki API.
Used by https://github.com/Tomionkkas/edith for two things: training the
EDITH model, and building the BM25 retrieval index the terminal answers from.
The index is not distributed - it is a pickle, unpickling executes arbitrary
code, and it rebuilds from this corpus in… See the full description on the dataset page: https://huggingface.co/datasets/Tomionkkas/edith-marvel-corpus.edit_amazon_reviews_multi_es
Dataset Summary
The data file is intended for a tutorial: Summarization
Language
Spanish
Dataset Structure
id: record id
stars: An int between 1-5 indicating the number of stars.
review_body: The text body of the review.
review_title: The text title of the review.
language: The string identifier of the review language.
product_category: String representation of the product's category.
lenght_review_body: text length of review_body
lenght_review_title: text… See the full description on the dataset page: https://huggingface.co/datasets/KRadim/edit_amazon_reviews_multi_es.code-edit-quality
Code Editing Quality — SFT-Ready (ShareGPT Format)
Quality-filtered splits of a 50K code-editing SFT dataset in ShareGPT conversation format, produced by LLM-based distillation that evaluates 9 quality criteria per sample.
Format
Each sample has a conversations field with ShareGPT-style turns:
system: Code editing system prompt
human: Instruction + source code
gpt: Edited code
Compatible with axolotl, LLaMA-Factory, and other SFT frameworks that support ShareGPT format.… See the full description on the dataset page: https://huggingface.co/datasets/AadiBhatia/code-edit-quality.reolyy-edit-intent-to-timeline
Reolyy Edit Intent to Timeline
Dataset Description
Creator intent plus clip metadata and transcript slices mapped to structured timeline edits.
Team Attribution
This dataset was created and reviewed by the Zarnite team through internal benchmark design, generation, and quality-control workflows. It should be presented as a Zarnite-authored benchmark starter pack, not as a purely human-collected field corpus.
Ecosystem Need Tier
High Ecosystem Need… See the full description on the dataset page: https://huggingface.co/datasets/zarnite/reolyy-edit-intent-to-timeline.wiki-editsThe pre-training dataset of paper "G-SPEED: General SParse Efficient Editing MoDel".
Visit https://github.com/Banner-Z/G-SPEED.git for more details.
fusion-aya-math-bench
Dataset Card for Fusion Aya Math Bench
Summary
Fusion Aya Math Bench is a multilingual, olympiad-level mathematical reasoning dataset. Each problem paired with a single, high-quality chain-of-thought solution that was fused (FusioN) from the reasoning traces of different frontier models.
Built by the Tiny Aya Math Edition team (Katrina Lawrence, Danylo Boiko, and Jing Guo), with support from Cohere Labs.
Pipeline
Derived from the open-ended… See the full description on the dataset page: https://huggingface.co/datasets/tiny-aya-math-edition/fusion-aya-math-bench.code-edit-dpo
Description
DPO preference pairs for code editing: prompt, chosen, rejected, plus the rule/reason the rejected answer violates.
Derived dataset. Source material retains its original per-item licence (see source/repo columns); treat as other / mixed. Provided as-is.
Usage
from datasets import load_dataset
ds = load_dataset("PotatoHD/code-edit-dpo")
Africans-THE-HISTORY-OF-A-CONTINENT-Second-Edition
Africans-THE-HISTORY-OF-A-CONTINENT-Second-Edition
Dataset created with PDF2Dataset -- OCR + structure-aware chunking pipeline.
Dataset Summary
Metric
Value
Total chunks
1500
Avg chars/chunk
718
Avg images/chunk
0.01
Source files
1
Duplicates removed
14
Quality filtered
88
Schema
Column
Type
Description
chunk_id
string
Unique identifier: filename_chunk_N
text
string
Raw markdown chunk with image refs… See the full description on the dataset page: https://huggingface.co/datasets/Svngoku/Africans-THE-HISTORY-OF-A-CONTINENT-Second-Edition.qwen3-hermes-strict-toolcall-synthetic-v4
Qwen3 Hermes Strict Tool-Call Synthetic V4
Registry status
Registry ID: edithatogo/qwen3-hermes-strict-toolcall-synthetic-v4
Family: hermes
Repository role: canonical_training_dataset
Canonical dataset: edithatogo/qwen3-hermes-strict-toolcall-synthetic-v4
Operational status: active
Rights status: apache-2.0-synthetic
Authoritative catalog: edithatogo/dataset-estate-registry
Origin and provenance
Origin repository:… See the full description on the dataset page: https://huggingface.co/datasets/edithatogo/qwen3-hermes-strict-toolcall-synthetic-v4.edit_amazon_reviews_multi_en
Dataset Summary
The data file is intended for a tutorial: Summarization
Language
English
Dataset Structure
id: record id
stars: An int between 1-5 indicating the number of stars.
review_body: The text body of the review.
review_title: The text title of the review.
language: The string identifier of the review language.
product_category: String representation of the product's category.
lenght_review_body: text length of review_body
lenght_review_title: text… See the full description on the dataset page: https://huggingface.co/datasets/KRadim/edit_amazon_reviews_multi_en.mol_edit_data
MolAct-Instruct Dataset
This dataset is used to train MolAct, an Agentic RL framework for molecular editing and optimization.
Description
The dataset is derived from ChemCoTBench. We extracted the source molecules (SMILES) and task specifications (editing instructions or optimization objectives) while removing the intermediate Chain-of-Thought (CoT) reasoning steps to fit the Reinforcement Learning environment.
Stage 1 (Editing): Focuses on functional group addition… See the full description on the dataset page: https://huggingface.co/datasets/little1d/mol_edit_data.HelpSteer3-good_edited_responseMake dataset from https://huggingface.co/datasets/nvidia/HelpSteer3 (good_edited_response)
who-covid-19-epidemiological-update-edition-163
World Health Organization (WHO) Epidemiological Update - Edition 163 (for embeddings)
train.pnf is taken from the WHO website
test.csv was generated by GPT-3.5-turbo
All text is chunked to a length of 500 tokens with 10% overlap.
arXivEdits_edits
Dataset Card for ArXivEdits (Edits)
ArXivEdits is a dataset comprising 751 English scientific papers from arXiv, each with sentence alignments across multiple revisions.
It also includes fine-grained, span-level edits which are annotated with the revision type and the underlying intention for 1000 sentences.
This dataset consists of only the edits subset of the whole dataset. The sentence-aligned papers can be found in this dataset.
Dataset Sources
Check the… See the full description on the dataset page: https://huggingface.co/datasets/miwytt/arXivEdits_edits.Trendyol-Cybersecurity-Instruction-Tuning-Dataset-EDIT
Trendyol Cybersecurity Defense Instruction-Tuning Dataset (v2.0)
🚀 TL;DR
53,202 meticulously curated system/user/assistant instruction-tuning examples covering 200+ specialized cybersecurity domains. Built by the Trendyol Security Team for training state-of-the-art defensive security AI assistants. Expanded from 21K to 53K rows with comprehensive coverage of modern security challenges including cloud-native threats, AI/ML security, quantum computing risks… See the full description on the dataset page: https://huggingface.co/datasets/JEAPI/Trendyol-Cybersecurity-Instruction-Tuning-Dataset-EDIT.Trendyol-Cybersecurity-Instruction-Tuning-Dataset-EDIT
Trendyol Cybersecurity Defense Instruction-Tuning Dataset (v2.0)
🚀 TL;DR
53,202 meticulously curated system/user/assistant instruction-tuning examples covering 200+ specialized cybersecurity domains. Built by the Trendyol Security Team for training state-of-the-art defensive security AI assistants. Expanded from 21K to 53K rows with comprehensive coverage of modern security challenges including cloud-native threats, AI/ML security, quantum computing risks… See the full description on the dataset page: https://huggingface.co/datasets/invinciblejha01/Trendyol-Cybersecurity-Instruction-Tuning-Dataset-EDIT.STEM_train_edit_predictions_test
STEM Image Edit Predictions Dataset
This dataset contains AI-generated edit predictions for STEM images based on captions and edit commands.
Dataset Structure
The dataset is organized in batches:
Total batches: 1
Each batch is stored in a separate directory (batch_0000, batch_0001, etc.)
Fields
Each item contains:
layout_summary: Concise description of source image layout and key elements
edit_analysis: Specific elements to be edited and changes to be made… See the full description on the dataset page: https://huggingface.co/datasets/JackyZhuo/STEM_train_edit_predictions_test.