Multilingual triplets for KaLM embedding training
From 1 to 10M examples with positive and negative pairs in English, Chinese, and other languages for embedding model fine-tuning.
- Embeddings
- English
- Chinese
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
From 1 to 10M examples with positive and negative pairs in English, Chinese, and other languages for embedding model fine-tuning.
An SQLite database of Danbooru2023 tags and posts with a latest ID of 7,866,491 for generative models.
English dataset with complex queries and texts of varying lengths used to train the ReasonIR-8B retriever.
A multimodal benchmark of H&E patches and gene expression profiles for image classification tasks.
A cleaned corpus of official public Indian documents in parquet format for search and question-answering systems.
10 million video clips with automatically generated captions and UMT scores for video-text tasks.
Dataset of 20,000 invoice images with named entity recognition markup and JSON annotations.
English text corpus of approximately 1.6B tokens covering red-team and blue-team topics.
About 143 million Chinese image-text pairs from LAION-2B for multimodal tasks.
A large-scale Vietnamese corpus for information retrieval and text ranking tasks.
An archive of all questions and answers from the Law section of StackExchange in English up to August 2023.
Tabular and text corpus in English for RLHF, response evaluation, and model policy compliance checking.
10k to 100k general knowledge and reasoning question-answer pairs.
An experimental English and Urdu bilingual corpus for MLM pre-training and contrastive embedding tuning.
8,094 IPCC report statements with expert annotation for evaluating LLM climate knowledge.
Bhagavad Gita verses in Sanskrit with Hindi and English translations for translation and generation tasks.
Tabular dataset on connections and co-occurrences of Marvel comic book heroes.
A French dataset labeled by LLaMA-3-70B for training embeddings and legal QA.
A high-resolution collection of 20th-century analog photos, with detailed descriptions already generated for 20k of them.
A CNN news corpus for simultaneous training and evaluation of summarization and QA retrieval.
Ranking of 8,125 GitHub developers for 2015-2025 based on Wayback Machine data.
Over 21,000 fruit images with pre-split train/test sets and CSV annotations.
A small set of semi-structured English cocktail recipes for text generation.
A corpus of over 35,000 patches for security research and vulnerability patching.