Which term is defined as 'occurs when the model outputs are not relevant or accurate'?

• A. Temperature
• B. False Positive
• C. Persona
• D. Input Content

Answer: (A)

Outputs that aren’t relevant or accurate reflect a reliability issue in language models, often described as hallucination or non-factual generation. The temperature setting is the knob that controls how random the model’s replies are. When temperature is high, the model explores a wider range of possibilities, which can lead to more diverse but less focused and sometimes irrelevant or incorrect content. When temperature is low, the output tends to be more deterministic and on-topic, usually improving relevance and accuracy. The other terms refer to different ideas: false positives involve misclassifying negatives as positives, persona is about adopting a specific character or voice, and input content is simply what you feed into the model.