21 terms
Prompting & Context
Talking to models well: prompts, examples, retrieval, memory, and managing the context window.
- chain of thoughtnountrendinga series of intermediate reasoning steps a model writes out before its final answer, either because it was prompted to or because…
- chunkingnounsplitting long documents into smaller pieces before creating embeddings, so a retrieval system can find and return just the…
- citationnounin AI answers, a link or reference showing which source supports a statement
- context engineeringnountrendingthe practice of deciding exactly what information, instructions, tools, and history go into a model's context window at each…
- context rotnounthe gradual decline in a model's accuracy and focus as its context window fills with more and more text, even before reaching the…
- few-shotadjectivedone by including a handful of examples in the prompt so the model can copy the pattern, without any retraining
- groundingnounconnecting a model's answer to trusted sources, such as documents, databases, or search results, so its claims can be checked and…
- in-context learningnouna model's ability to pick up a new task from instructions or examples given in the prompt, without changing its weights
- memorynountrendinga feature that lets an AI keep information about a user or task across separate conversations, such as preferences, past…
- one-shotadjectivetrendingdone with exactly one example included in the prompt
- personanouna role or character an AI is told to adopt, such as 'experienced editor' or 'friendly tutor', which shapes its tone and focus
- promptnountrendingthe text, question, or instruction a person gives to an AI model to get a response
- prompt cachingnounreusing the already-processed beginning of a prompt across many requests, which cuts cost and speeds up responses when the same…
- prompt engineeringnounthe practice of designing and refining prompts to get reliable, high-quality results from an AI model
- rerankingnouna second step in search where a more careful model re-scores the top results so the most relevant ones come first
- retrieval-augmented generationnountrendinga technique in which a system first searches a collection of documents for relevant passages and then gives them to the model, so…
- semantic searchnounsearch that matches the meaning of a query rather than its exact words, usually by comparing embeddings
- structured outputnounmodel output that follows a fixed, machine-readable format such as JSON with specific fields, so software can use it directly
- system promptnounhidden instructions given to a model before the conversation starts, which set its role, tone, rules, and limits for every reply
- vector databasenouna database built to store embeddings and quickly find the ones most similar to a query, used for semantic search and RAG
- zero-shotadjectivedone without giving the model any examples, relying only on instructions
Advertisement