Collection of jailbreak prompts for ChatGPT
A small set of ChatGPT jailbreak prompts for studying constraint bypassing.
- Question answering
- Text generation
- Fill mask
- English
- Afar
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 small set of ChatGPT jailbreak prompts for studying constraint bypassing.
A selection of real, complex English prompts from WildChat for evaluating large language models.
About 6 million question-answer pairs in 21 languages collected from the internet for search and QA systems.
7,469 test questions from South Korean medical licensing exams.
A collection of 21,318 prompts and attacks (jailbreaks, toxicity) for testing and improving neural network safety.
Educational subset from a massive corpus of translations between English and over 500 languages.
Large structured archive of Indian judicial acts, tribunal decisions, and legislation in English.
A database of 34,831 computer science research papers with section splits and ready-to-use fine-tuning data.
Combined table of Python questions, answers, and tags from Stack Overflow for 2008–2016.
Massive corpus of over 140,000 public domain English books with OCR texts.
A large Russian instruction-response dataset for training models to follow commands.
A prompt set with 70 real-life scenarios for testing LLMs for age and gender bias.
Evaluation sets for LLM domain adaptation in finance from the ICLR 2024 paper.
Cleaned logs of multi-step operation by the OpenAI Codex Desktop agent in a local Windows environment.
Tabular and text corpus for training and evaluating language model safety using RLHF.
An English set of 177,597 financial instructions and Q&A pairs for LLM fine-tuning.
Over 5 million sections of US laws, regulations, and prosecutor opinions for search and analysis.
200k Chinese instruction-response pairs for aligning and training language models.
A large multi-task streaming audio dataset for training audio LLMs and voice agents.
Raw JSONL agent traces from Qwen3.7 Max with a ready-to-use tool schema.
Evaluation of linguistic world models using agent trajectories and ground-truth observations from real environments.
Synthetic SaaS vendor and client dialogues for conversion prediction and sales model training.
999,847 personas with 1,290 categorical attributes for character generation and analysis.
Cleaned code across 600+ programming languages, grouped by repository, for training LLMs.