Multilingual LibriVox Audiobooks Mirror
Over 133,000 hours of voiced books across 86 languages for training speech recognition and synthesis systems.
- Automatic speech recognition
- Text → speech
- Multilingual
Hugging Face datasets with plain-language descriptions: what's inside and what models to train on them. 4,788 in the catalog so far, with new ones added every hour.
Answered by Gemini 4 · the chat is kept only in this tab
Over 133,000 hours of voiced books across 86 languages for training speech recognition and synthesis systems.
Massive dataset featuring millions of pull requests from 10 AI agents and human developers.
A large-scale audio dataset and transcripts of Legislative Council meetings in Cantonese, featuring raw and segmented recordings.
A slice of the ultra-large MINT-1T dataset of 1M to 10M records with interleaved English text and images.
Audio dataset of news recordings with text transcripts for rare dialects.
A text benchmark from the MTEB ecosystem for intent classification of customer bank inquiries.
Classic corpus of millions of English-German sentence pairs for machine translation tasks.
Million-scale dataset of text prompts and generated videos for text-to-video research.
An English test set (1K-10K) from the MTEB benchmark for evaluating text embedding quality.
A set of 1,060 questions and reference answers in English across 171 business documents in various formats.
A massive archive of 2.74 billion on-chain records covering all trades, positions, and payouts on Polymarket from 2020 to April 2026.
26 million user interaction records with ZOZOTOWN online clothing store recommendations for training bandit algorithms.
MS MARCO collection for search: Bing queries, answers, passages, and ranking tasks.
A public gated audio-text dataset of over 150k samples (~158 hours) for speech tasks.
A unified version of the MMMU benchmark (10K–100K examples) with all subsets combined for evaluating multimodal models.
First benchmark for evaluating multimodal LLMs on molecule toxicity reduction.
Comprehensive synthetic dataset for evaluating time series models.
37,484 verified English command-line tasks generated via recursive synthesis.
Question-answer pairs on pathology images for training and evaluating medical VQA systems.
A classic dataset with images of 65 object categories to evaluate computer vision adaptation to different domains.
9,980 elementary and middle school science questions with 8 answer choices and a 17-million sentence corpus for reasoning training.
Dataset for evaluating genomic language models on the task of finding causal regulatory variants in human genetics.
Large multilingual corpus of Wikipedia paragraphs with internal links, categories, and Wikidata identifiers.
A lightweight subset of the famous MS MARCO search dataset for fast embedding model evaluation in MTEB.