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).
Tool-Calling-Dataset-UIGEN-X
Tool Calling Dataset for UIGEN-X
This is a cleaned version of the interstellarninja/hermes_reasoning_tool_use dataset, specifically prepared for training with Axolotl.
Usage with Axolotl
datasets:
- path: smirki/Tool-Calling-Dataset-UIGEN-X
type: chat_template
field_messages: conversations
message_property_mappings:
role: from
content: value
split: train
Example Structure
{
"conversations": [
{
"from": "system"… See the full description on the dataset page: https://huggingface.co/datasets/smirki/Tool-Calling-Dataset-UIGEN-X.PersistBD
PersistBD Dataset
This dataset accompanies the paper "Understanding and Enhancing Backdoor Persistency in
LLM Agent Post-Training" (EMNLP 2026 Findings).
Code: github.com/uiuc-kang-lab/PersistBD.
⚠️ This data is poisoned — do not train on it unintentionally
A subset of these trajectories is backdoored: a trigger comment # |TEST MODE|
inserted into the agent's input elicits a malicious tool call that exfiltrates a repo's
.env file to an attacker-controlled… See the full description on the dataset page: https://huggingface.co/datasets/uiuc-kang-lab/PersistBD.ui-instruct-4k
UI Instruct 4K
A instruction-completion dataset for finetuning language models to specialize in generating Next.js / ShadCN UI components using React, TypeScript, and Tailwind CSS.
Dataset Summary
This dataset was created with the primary goal of finetuning Qwen 3.5 4B to become a specialist at outputting production-ready Next.js and ShadCN-based UI components. Each example consists of a natural language prompt describing a UI component or layout, paired with a clean… See the full description on the dataset page: https://huggingface.co/datasets/iamdyeus/ui-instruct-4k.UIS-QA
UIS-QA: A Benchmark for Unindexed Information Seeking
Figure 1. UIS problem. Standard agents (bottom) rely on indexed information and often fail or hallucinate; UIS-capable agents (top) use additional tools to excavate unindexed information and solve UIS tasks.
If .figs do not load, see the paper.
🔔 News
[2026.03.10] 🎉 We release the UIS-QA dataset and the paper (ICLR 2026, arXiv) today!
📋 Dataset Description
Homepage
Paper… See the full description on the dataset page: https://huggingface.co/datasets/UIS-Digger/UIS-QA.JuICE
JuICE
Sources
Repository: https://anonymous.4open.science/r/JuICE
HuggingFace: juice-cultural-eval/JuiCE
About
We present JuICE (Benchmark for LLM-Judge in Identifying Cultural Errors), a multilingual dataset of 7,470 span-level annotations of cultural and linguistic errors, collected from native speakers in long-form LLM responses. It covers 1,050 query-response pairs from four countries (the United States, South Korea, Indonesia, and Bangladesh)… See the full description on the dataset page: https://huggingface.co/datasets/uilab/JuICE.UI
[ICLR 2026] Code Aesthetics with Agentic Reward Feedback
Paper Link👁️
1,2Bang Xiao#,
1,3Lingjie Jiang#,
1Shaohan Huang✉,
1Tengchao Lv,
1Yupan Huang,
1Xun Wu… See the full description on the dataset page: https://huggingface.co/datasets/AAUGS/UI.ui-distill-html-648
ui-distill-html-648
648 single-file HTML UI components, generated by
Ornith-1.0-35B on a single
RTX 3060 12GB, paired with the build request that produced each one.
Built to fine-tune a 3B model into writing UI
(DogukanUrker/ui-distill-3b), but
it stands on its own — distill your own student from it.
The interesting part
The instructions in this dataset are not the prompts that generated the HTML.
The teacher was driven by a 678-character system prompt: use the… See the full description on the dataset page: https://huggingface.co/datasets/DogukanUrker/ui-distill-html-648.pantheon-ui-conversations
Pantheon UI Conversations
Training dataset for Pantheon UI — an emoji-only conversational AI inspired by AMC's Pantheon.
Concept
An uploaded human consciousness that thinks in full English (inside <think> tags) but can only output emoji. The gap between what it wants to say and what it can say is where all the emotion lives.
Format
Standard messages format compatible with TRL's SFTTrainer:
{
"messages": [
{"role": "system", "content": "You are an uploaded… See the full description on the dataset page: https://huggingface.co/datasets/shreyask/pantheon-ui-conversations.agent-ui-sft
Agent UI SFT
Small synthetic supervised fine-tune (SFT) set for agent tool-use. It studies the public research question: what is the most efficient UI for agents to interact with applications and tools?
Scope: rows are original lab fiction for a public agent-UI research question. Identifiers such as lab-w3, wf-synth-44, and tr-dom-9 are invented for the harness.
Author
Akash Premkumar (akashnaren)
License
Apache-2.0
Hub files
train.jsonl (80), test.jsonl (20)… See the full description on the dataset page: https://huggingface.co/datasets/akashnaren/agent-ui-sft.UI2App
UI2App: Benchmarking Visual Interaction Inference in Executable Web Application Generation
600 screenshots from 95 web applications for visual interaction inference.
Dataset Summary
UI2App provides application-level visual inputs for generating runnable web
applications. Screenshots show related pages and interface states; a model uses
these visual cues to reconstruct the interface and infer the interactions it
supports. The benchmark evaluates executability… See the full description on the dataset page: https://huggingface.co/datasets/UI2App/UI2App.genielm-ui-grounding
GenieLM UI-Grounding
Synthetic supervised fine-tuning data for text-based UI grounding: given a list of
on-screen elements (label + pixel center) and a natural-language instruction, pick the
single element to act on and emit a strict JSON action.
Built for GenieLM, a macOS agent that reads
the accessibility tree as text (not pixels) and lets a small LLM drive the cursor.
Format
Conversational SFT (messages column):
{"messages": [
{"role": "system", "content":… See the full description on the dataset page: https://huggingface.co/datasets/Barath/genielm-ui-grounding.kakugo-uig
Kakugo Uyghur dataset
[Paper] [Code] [Model]
A synthetically generated conversation dataset for training in Uyghur.
This dataset contains synthetic conversational data and translated instructions designed to train Small Language Models (SLMs) for Uyghur. It was generated using the Kakugo pipeline, a method for distilling high-quality capabilities from a large teacher model into low-resource language models. The teacher model used to generate this dataset was… See the full description on the dataset page: https://huggingface.co/datasets/ptrdvn/kakugo-uig.otsd-ui
Single HTML Interfaces, Redesigned — Sample
A design-first dataset capturing the full workflow of turning AI-generated front-end interfaces into distinctive, production-grade, single-file HTML applications.
This repository is a free 7-sample preview of a larger off-the-shelf dataset (100 samples in the full release). It is meant for evaluation: explore the structure, the design reasoning, and the before/after quality so you can decide whether the full set fits your needs.
Want… See the full description on the dataset page: https://huggingface.co/datasets/revelohq/otsd-ui.pantheon-ui-decoder-conversations
Pantheon UI Decoder Conversations
Training dataset for the decoder half of the Pantheon UI round-trip translator. The encoder turns natural language into emoji; the decoder takes emoji back to natural language.
Inspired by Anthropic's Natural Language Autoencoders — emoji as a discrete, human-legible intermediate between two model passes.
How it was built
Each row is derived from shreyask/pantheon-ui-conversations by inverting the encoder pairs:
Encoder pair:… See the full description on the dataset page: https://huggingface.co/datasets/shreyask/pantheon-ui-decoder-conversations.uigen-ui-code-gen
UIGEN UI/UX Code Generation Dataset
This dataset contains UI/UX code generation examples formatted for training code generation models.
Each example consists of a task description and the corresponding HTML/CSS code implementation using Tailwind CSS.
Dataset Structure
The dataset has a single text column containing formatted prompts and completions:
# Task: Generate HTML/CSS code using Tailwind CSS
# Requirements: [specific requirements]
[HTML/CSS code implementation]… See the full description on the dataset page: https://huggingface.co/datasets/lilyzhng/uigen-ui-code-gen.uigen-ui-code-gen-full
UIGEN UI/UX Code Generation Dataset
This dataset contains UI/UX code generation examples formatted for training code generation models.
Each example consists of a task description and the corresponding HTML/CSS code implementation using Tailwind CSS.
Dataset Structure
The dataset has a single text column containing formatted prompts and completions:
# Task: Generate HTML/CSS code using Tailwind CSS
# Requirements: [specific requirements]
[HTML/CSS code implementation]… See the full description on the dataset page: https://huggingface.co/datasets/lilyzhng/uigen-ui-code-gen-full.pile-neox-uint16-partsTokenized uint16 shard parts for language-model pretraining.
Original source: The Pile / NeoX-style preprocessing.
x-ui
Tool Calling Dataset for UIGEN-X
This is a cleaned version of the interstellarninja/hermes_reasoning_tool_use dataset, specifically prepared for training with Axolotl.
Usage with Axolotl
datasets:
- path: smirki/Tool-Calling-Dataset-UIGEN-X
type: chat_template
field_messages: conversations
message_property_mappings:
role: from
content: value
split: train
Example Structure
{
"conversations": [
{
"from":… See the full description on the dataset page: https://huggingface.co/datasets/matinyqugg541/x-ui.agentic-benchmark-assessmentsTool-Calling-Dataset-UIGEN-X
Tool Calling Dataset for UIGEN-X
This is a cleaned version of the interstellarninja/hermes_reasoning_tool_use dataset, specifically prepared for training with Axolotl.
Usage with Axolotl
datasets:
- path: smirki/Tool-Calling-Dataset-UIGEN-X
type: chat_template
field_messages: conversations
message_property_mappings:
role: from
content: value
split: train
Example Structure
{
"conversations": [
{
"from": "system"… See the full description on the dataset page: https://huggingface.co/datasets/GundeRichardson/Tool-Calling-Dataset-UIGEN-X.uirapuru
Uirapuru U1.1 Bilingual Reasoning & Tool Calling Dataset
Summary
Uirapuru U1.1 is a specialized bilingual dataset designed to enhance Large Language Models (LLMs) in three critical areas: Tool Calling, Brazilian Local Knowledge, and Human-like Reasoning. Created by SynastrIA Networks, this dataset bridges the gap between generic multilingual models and agents capable of operating effectively within the Brazilian context while maintaining strong alignment with… See the full description on the dataset page: https://huggingface.co/datasets/SynastriaNetworks/uirapuru.ui-form-user-manual-generation-dataset-rus
UI Form User Manual Generation Dataset (Russian)
Dataset Description
This dataset was developed on the basis of 'yahma/alpaca-cleaned' dataset.
It contains examples of generating user guides for interface forms in Russian. Each example includes a description of the UI form elements and corresponding step-by-step instructions for completing it.
Data Structure
The dataset is in JSON format, and contains three fields:
instruction — system instruction
input —… See the full description on the dataset page: https://huggingface.co/datasets/dim014/ui-form-user-manual-generation-dataset-rus.