AI Concepts category
Prompting & Generation
Chain-of-Thought
Chain-of-thought is intermediate reasoning text generated as part of producing an answer, often elicited through prompting.
Constrained Decoding
Constrained decoding limits token choices during generation so output follows a specified grammar or schema.
Context Window
A context window is the bounded amount of input and generated tokens a model can process in one request.
Few-Shot Prompting
Few-shot prompting supplies a small set of worked examples in the model input to guide how it handles a new case.
In-Context Learning
In-context learning is adapting model behavior through instructions or examples in the current request without updating model weights.
Inference-Time Compute
Inference-time compute is the computation spent while generating and selecting an answer after model training.
Prompt Chaining
Prompt chaining passes model output into a later prompt as part of a predefined workflow.
Sampling
Sampling draws each output token according to its probability, allowing different continuations from the same prompt.
Stopping Condition
A stopping condition is a rule that ends a model's generation, whether it finishes naturally or reaches a configured boundary.
Structured Outputs
Structured outputs constrain model responses to a machine-readable structure, often specified with JSON Schema.
System Prompt
System prompts shape model behavior through instructions or context supplied apart from user input.
Temperature
Temperature adjusts the sharpness of the token probability distribution used during sampling.
Top-k Sampling
Top-k sampling restricts token selection to the fixed number of highest-probability candidates at each generation step.
Top-p Sampling
Top-p sampling restricts token selection to the smallest set of highest-probability tokens whose cumulative probability reaches a threshold.
Zero-Shot Prompting
Zero-shot prompting asks a model to perform a task without including worked examples in the prompt.