What is the key feature of the Large Language Models (LLMs)?

What is the key feature of the Large Language Models (LLMs)?

• A. Classifying sentiment in text
• B. Translating between languages
• C. Predicting the next word (formally called a 'token')
• D. Generating images from prompts

Answer: (C)

The key feature of Large Language Models is predicting the next word (token) given the preceding text. They are trained to assign high probability to the next token in a sequence, learning how language tends to unfold by seeing vast amounts of text. When you provide a prompt, the model generates text by repeatedly selecting the most probable next token, producing coherent and contextually relevant output. A token is typically a word or a subword piece, and the training objective is to maximize the likelihood of the next token conditioned on what came before. This autoregressive prediction is what enables not only fluent generation but also the model’s versatility across many tasks when prompted, even if sentiment analysis or translation can be achieved as downstream uses. Generating images falls outside this text-focused framework, which is why that option isn’t the defining feature.

The key feature of Large Language Models is predicting the next word (token) given the preceding text. They are trained to assign high probability to the next token in a sequence, learning how language tends to unfold by seeing vast amounts of text. When you provide a prompt, the model generates text by repeatedly selecting the most probable next token, producing coherent and contextually relevant output. A token is typically a word or a subword piece, and the training objective is to maximize the likelihood of the next token conditioned on what came before. This autoregressive prediction is what enables not only fluent generation but also the model’s versatility across many tasks when prompted, even if sentiment analysis or translation can be achieved as downstream uses. Generating images falls outside this text-focused framework, which is why that option isn’t the defining feature.