What are LLMs mainly trained on?
AnswerLarge amounts of text
They learn patterns of language from huge text collections.
AI · Facts & stats
LLM facts & stats
7 fact-checked facts about large language models and how they work, each with the reason behind it. 5 more are in today's round and join this page after it closes.
The basics
In AI tools, what is a prompt?
AnswerThe input you give the model
A prompt is the text or instruction the model responds to.
Going deeper
In LLMs, what are tokens?
AnswerChunks of text the model processes
Text is split into tokens, which may be words or word pieces.
What does GPT stand for?
AnswerGenerative Pre-trained Transformer
GPT models are pre-trained on text and generate new text.
What does RLHF stand for?
AnswerReinforcement learning from human feedback
RLHF uses human ratings to steer model behavior.
Expert level
What does raising the temperature setting of an LLM generally do?
AnswerMakes output more random
Higher temperature spreads probability across more word choices.
What does the word2vec method produce?
AnswerWord embeddings (vectors)
Word2vec, from 2013, maps words to vectors that capture meaning.
Know these cold?Today's round is five questions like these, 20 seconds each. Your streak and level follow you to every arena.Play today's round