AI Dictionary › AI Fundamentals

Beam search

Beam search (ricerca a fascio)

Beam search is a decoding strategy that, instead of choosing one word at a time, keeps several candidate sequences alive in parallel to find the overall most probable sentence. Imagine exploring a maze while keeping the five most promising paths (the "beam") open at once: at each fork you extend all five, discard the worst, and open new ones. In the end you pick the route with the best total score, not the one that merely looked good at the first step. This avoids the trap of a greedy early choice leading to a dead end.

Definition

Beam search matters mostly in machine translation and tasks with a well-defined "correct" answer. In conversational models it is less used because it tends to produce generic, repetitive text, trading naturalness for probability.

Related terms

• Greedy decoding
• Sampling
• Top-p (nucleus sampling)
• Top-k
• Perplexity (metric)

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.