# AI Sampling Parameters Parameters that control how an LLM selects the next token during generation. Beyond temperature, the key parameters are: - **Top-p (nucleus sampling)**: only consider tokens whose cumulative probability reaches p (e.g., 0.9 = consider the smallest set of tokens that together have 90% probability) - **Top-k**: only consider the k most probable tokens - **Frequency penalty**: reduces probability of tokens that already appeared (prevents repetition) - **Presence penalty**: reduces probability of any token that appeared at all (encourages topic diversity) These parameters interact with each other and with temperature. Most APIs let you set combinations. ## References ## Related - [[Large Language Models (LLMs)]] - [[Prompt Engineering]]