AI Dictionary › AI Fundamentals

Sampling

Campionamento

Sampling is how a language model chooses the next word among many possibilities. At each step the model produces not a single certain answer but a probability distribution over candidate tokens: "cat" 40%, "dog" 25%, "horse" 10% and so on. Sampling is the act of drawing from this weighted bag. It is like rolling loaded dice where likely faces appear more often, but not always. Raising or lowering randomness yields more creative, unpredictable text or more cautious, repetitive text. Techniques like top-p, top-k and temperature are all variants of how this draw is performed.

Definition

Sampling matters because it sets the character of an output: two identical calls can differ. Tuning it well separates a dull assistant from a lively one, and reliability from hallucination.

Related terms

• Top-p (nucleus sampling)
• Top-k
• Temperature
• Greedy decoding
• Beam search

More in AI Fundamentals

• LLM (Large Language Model)
• Token
• Context Window
• Temperature
• AI Hallucination
• Fine-tuning
• RAG (Retrieval Augmented Generation)
• Embedding

Put it into practice

• Real scenario: apply this concept with an instant AI verdict.
• Daily gym: short drills to train dictionary concepts.
• Playoff game: challenge other players, elimination-style.

📱 Download the Android app (beta) iOS coming soon

Say what you mean. Get what you need.

Grace Certified, the AI coach that trains and certifies your prompt engineering, by Agora Intelligence.