Sampling
Understand sampling in language generation, with temperature, top-k, top-p, and the limits of reproducibility.
-
-
Decoding
Learn how generation decoding selects tokens from model scores, and distinguish it from converting token IDs back to text.
-
Context Window
Understand context capacity, output budgets, and why conversation storage does not guarantee that a model sees every detail.
-
Prompt
Understand a prompt as request-time input and distinguish it from model training, stored memory, and access control.
-
Tokenization
Understand how text becomes model input units and why normalization, vocabulary, and model compatibility matter.
-
Token
Learn what a text token represents, why it differs from a word, and how token counts depend on the tokenizer.
-
Generative AI
Understand generative AI, its relationship to language models, and why generated content still needs evidence and checks.
-
Retrieval-Augmented Generation (RAG)
Learn how RAG uses retrieved context for generation, how it differs from training, and where errors can enter.