datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
MathInstruct
🦣 MAmmoTH: Building Math Generalist Models through Hybrid Instruction Tuning
MathInstruct is a meticulously curated instruction tuning dataset that is lightweight yet generalizable. MathInstruct is compiled from 13 math rationale datasets, six of which are newly curated by this work. It uniquely focuses on the hybrid use of chain-of-thought (CoT) and program-of-thought (PoT) rationales, and ensures extensive coverage of diverse mathematical fields.
Project Page:… See the full description on the dataset page: https://huggingface.co/datasets/TIGER-Lab/MathInstruct.FIM-Midtraining-400K
FIM-Midtraining-400K
📄 Paper · 💻 GitHub · 🤗 Collection
The mid-training corpus of "Function-Aware Fill-in-the-Middle as Mid-Training for Coding Agent Foundation Models": 400K function-aware FIM samples (~2.6B tokens under the Qwen2.5-Coder tokenizer) drawn from 75,568 Python files across 968 permissively-licensed GitHub repositories, fully decontaminated against SWE-Bench.
A coding agent's inner loop — act → observe → continue — is structurally isomorphic to a function call… See the full description on the dataset page: https://huggingface.co/datasets/TIGER-Lab/FIM-Midtraining-400K.SWE-Next
SWE-Next: Scalable Real-World Software Engineering Tasks for Agents
SWE-Next Dataset
SWE-Next is an execution-grounded dataset of 2,308 self-verifying software engineering tasks mined from real merged GitHub pull requests. Starting from 3,971 seeded Python repositories and 102,582 executed candidate base/merged commit pairs, SWE-Next retains only instances where the merged commit produces a strict test improvement without regressions. The final release… See the full description on the dataset page: https://huggingface.co/datasets/TIGER-Lab/SWE-Next.SWE-Next-SFT-Trajectories
SWE-Next: Scalable Real-World Software Engineering Tasks for Agents
SWE-Next SFT Trajectories
SWE-Next SFT Trajectories is the supervised fine-tuning dataset released with SWE-Next: Scalable Real-World Software Engineering Tasks for Agents. It contains 3,693 ShareGPT-style multi-turn training examples collected from expert agent rollouts on 2,308 execution-grounded SWE tasks synthesized from real merged pull requests.
The dataset is designed for training… See the full description on the dataset page: https://huggingface.co/datasets/TIGER-Lab/SWE-Next-SFT-Trajectories.MetricInstruct
MetricInstruct
The MetricInstrcut dataset consists of 44K quadruple in the form of (instruction, input, system output, error analysis) for 6 text generation tasks and 22 text generation datasets. The dataset is used to fine-tune TIGERScore, a Trained metric that follows Instruction Guidance to perform Explainable, and Reference-free evaluation over a wide spectrum of text generation tasks.
Project Page | Paper | Code | Demo |
TIGERScore-7B | TIGERScore-13B
We present the… See the full description on the dataset page: https://huggingface.co/datasets/TIGER-Lab/MetricInstruct.BrowserAgent-Data
BrowserAgent ChatML Dataset (SFT/RFT)
This dataset contains ChatML-style multi-turn dialogues for a browser agent task. The data is prepared as JSON Lines so it can be previewed directly with the Hugging Face Hub Data Visualizer and loaded with the datasets library.
Links
Paper
Github
Files
sft.jsonl — SFT split (one JSON object per line)
rft.jsonl — RFT split (one JSON object per line)
Schema
Each record is a JSON object containing:
messages:… See the full description on the dataset page: https://huggingface.co/datasets/TIGER-Lab/BrowserAgent-Data.SVAMPSKGInstruct-skg-only
🏗️ StructLM: Towards Building Generalist Models for Structured Knowledge Grounding
SKGInstruct-skg-only is an instruction tuning dataset constructed from 19 structured knowledge grounding datasets.
Project Page: https://tiger-ai-lab.github.io/StructLM/
Paper: https://arxiv.org/pdf/2402.16671.pdf
Code: https://github.com/TIGER-AI-Lab/StructLM
Models:
7B | StructLM-7B
13B | StructLM-13B
34B | StructLM-34B
License
Dataset Name
License Type
TabMWP… See the full description on the dataset page: https://huggingface.co/datasets/TIGER-Lab/SKGInstruct-skg-only.infopedia-pt-ipa
European Portuguese IPA Lexicon — Infopédia
A lightweight word → IPA pronunciation lexicon for European Portuguese,
extracted from Infopédia (Porto Editora). One row
per headword, intended for grapheme-to-phoneme (G2P) work, pronunciation
modelling, and TTS/ASR lexicon building.
Complete crawl. Derived from a graph crawl of Infopédia that ran to
convergence (frontier → 0), covering the dictionary's reachable component.
Contents
Field
Count
Entries… See the full description on the dataset page: https://huggingface.co/datasets/TigreGotico/infopedia-pt-ipa.TigrinyaLargeText
TIgrinyaLargeTextDataset
Dataset Description
TIgrinyaLargeTextDataset is a large-scale collection of Tigrinya language articles designed for training Large Language Models (LLMs) and other Natural Language Processing (NLP) tasks. This dataset addresses the critical need for high-quality Tigrinya language resources in the machine learning community.
Dataset Summary
Language: Tigrinya
Task Categories: Language Modeling, Text Generation, NLP Research
Size: 12,374… See the full description on the dataset page: https://huggingface.co/datasets/mewaeltsegay/TigrinyaLargeText.haddas-tigrinya-corpus
haddas-tigrinya-corpus
Monolingual Tigrinya newspaper text segmented into article bodies. Suitable for continued pretraining or causal language modeling of Tigrinya LLMs.
Source
Derived from the Haddas Eritrea newspaper archive: 63 PDF issues processed
by the haddas-eritrea pipeline (extract -> clean -> segment -> translate -> label).
Generated: 2026-04-26 12:21 UTC
Row count: 2653
Schema: id, text, char_count, topic, issue_date, source_pdf, page_start, page_end… See the full description on the dataset page: https://huggingface.co/datasets/SIMBA9657/haddas-tigrinya-corpus.BrowerTrace-sample
BrowserTrace (Sample)
Overview
This repository hosts a sample slice of the BrowserTrace dataset: browser interaction trajectories captured via human annotators and post-processed into Chain-of-Thought (CoT) annotations. Each record captures one end-to-end task, the user-facing instructions, evaluator scores, and the step-by-step reasoning/actions taken by the agent. The JSONL file is paired with an archived set of screenshots.
Contents
data/train.jsonl: All… See the full description on the dataset page: https://huggingface.co/datasets/TigLabs/BrowerTrace-sample.