AI agent pressure resistance benchmark
A 10K–100K example test suite for checking whether an AI agent follows rules under user pressure.
- Question answering
- Text classification
- Text generation
- English
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
A 10K–100K example test suite for checking whether an AI agent follows rules under user pressure.
A text corpus from automotive subreddits for model training and analysis.
10k to 100k digitized multi-page National Archives documents in English.
433K sentence pairs with textual entailment annotations for training and evaluating NLI models.
Large DataComp-1B image-text corpus with improved captions from LLaVA-1.5-LLaMA3-8B.
An English dataset for predicting logical continuations of everyday situations to test model commonsense.
An English intent classification corpus that teaches dialog systems to recognize irrelevant queries.
Massive corpus of over 106 million English scientific articles and conference papers.
A text benchmark from the MTEB ecosystem for intent classification of customer bank inquiries.
Benchmark: the model receives an unstructured state and answers five typed questions with probabilities.
Large balanced dataset of 630K English texts for multi-class classification.
100K+ human-annotated QA pairs across 14 harm categories for AI safety.
Approximately 190 million Portuguese texts divided into European and Brazilian variants.
58K Reddit comments labeled across 27 emotions plus Neutral for classification.
Annotated English sentences for training models to find acronyms in documents.
Tens of millions of vectors for Wikidata elements in 4 languages from Wikimedia, Jina AI, and IBM.
Massive English corpus of 200M instruction-response pairs across 40+ task categories for multi-task model pre-training.
Combined corpus of 16 datasets with video, audio, and text for recognizing emotions, humor, and human state.
English-language corpus of current job postings directly from hiring system APIs for text search and classification tasks.
Largest open corpus for topic classification of Wikipedia sentences across 205 languages and dialects.
Millions of English comments from 50 news sites for training text classification and moderation systems.
Millions of classification, choice, rating, and visual reasoning examples for multi-task learning.
A massive corpus of company documents and reports from SEC EDGAR for training and evaluating financial language models.
176 languages, LOC/PER/ORG markup in IOB2 based on Wikipedia articles.