STS14 semantic textual similarity benchmark
The classic SemEval STS 2014 English dataset within MTEB for evaluating text embedding quality.
- Sentence similarity
- English
Hugging Face datasets with plain-language descriptions: what's inside and what models to train on them. 4,797 in the catalog so far, with new ones added every hour.
Answered by Gemini 4 · the chat is kept only in this tab
The classic SemEval STS 2014 English dataset within MTEB for evaluating text embedding quality.
Dataset for testing reward models on chats, complex dialogues, and preferences; English text only.
A multimodal corpus of trajectories and action reasoning for training robotic manipulator control models.
Between 10 and 100 million pre-processed Wikipedia texts for evaluating tokenization efficiency across 45 languages.
A dataset of images and 3D data from the 3D Arena project for testing 3D object generation.
A benchmark set of English prompts from LMSYS for testing conversational models.
A massive collection of Chinese image-text pairs totaling approximately 7.9 TB for image-to-text tasks.
English-language corpus of current job postings directly from hiring system APIs for text search and classification tasks.
A large-scale dataset of 10 to 100 million rows with various tabular and text features for evaluating tabular models.
Benchmark tasks for testing the ability of coding agents to independently conduct scientific research from data analysis to paper writing.
Official image-text dataset for the GeoText1652 research paper containing 100,000 to 1,000,000 records.
BBC news articles with single-sentence summaries for extreme summarization training.
100K English web snippets with metadata for checking neural network toxic degeneration.
Classic corpus of ~3.3 million image-text pairs collected from HTML Alt-text tags.
Table corpus with provenance metadata and context for training tabular models.
PsiBot multimodal dataset with vision, language, and actions for training embodied AI.
Question-answer pairs and executable reasoning traces on real Kaggle notebooks for data analysis.
Image and text dataset for evaluating basic visual perception skills in multimodal models.
A massive updated code dataset from TokyoTech-LLM for training program generation models.
Largest open corpus for topic classification of Wikipedia sentences across 205 languages and dialects.
2,323 scientific charts from arXiv papers to test multimodal model ability to understand complex visualizations.
Millions of English comments from 50 news sites for training text classification and moderation systems.
A set of text pairs from the MTEB benchmark to evaluate the quality of finding the best counter-argument without topic knowledge.
The complete corpus of arXiv LaTeX sources, parsed and aligned with official metadata in Parquet format.