AI Dictionary › AI Fundamentals

Perplexity (metric)

Perplessità

Perplexity is a metric measuring how "surprised" a language model is by the text it meets: the lower the value, the better it predicts the next word. Imagine a reader guessing a hidden word in a sentence: if few valid options come to mind and they are almost always right, perplexity is low; if they grope among dozens of possibilities, it is high. Technically it represents the average number of choices the model wavers between at each step. A model with perplexity 10 hesitates as if choosing among ten equally likely words; with perplexity 2, like flipping a coin.

Definition

Perplexity matters as a fast tool to compare models or versions during training. Its limit is that a good score does not guarantee useful, truthful or pleasant answers: it measures prediction, not perceived quality.

Related terms

• Benchmark
• Tokenizer
• Sampling
• Machine Learning
• Token

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.

Say what you mean. Get what you need.

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