liveBook · Manning

Temperature

As explained in Effective Conversational AI

Temperature is a generation parameter that controls the degree of randomness in a language model's token selection.

When a chatbot needs varied wording rather than the same response every time, temperature adjusts how much randomness enters token selection. Higher temperature generally produces more creative and diverse output, while greedy decoding remains deterministic. section 7.2.1 section 4.1.4

Temperature is useful for exploratory generation, such as expanding “I forgot my password” into alternative nouns and verb phrases for intent testing. It can produce useful variations, but also odd or unsuitable phrases, so generated results still require review. section 7.2.1

Creative synonym generation

With increased creativity settings, a model generated alternatives such as “PIN,” “access code,” and “security code” for “password,” alongside less reliable phrases such as “unknown,” “not applicable,” and “OMG failed to remember.” The broader list helps designers test more user wording, but should not be treated as an automatically correct dataset. section 7.2.1

Listing 7.5 Generating noun synonyms with increased creativity settings
You are a consumer contacting an online retailer via their chat service.

Generate a numbered list of synonyms a consumer might  #1
use for 'password' while they described a problem like   #1
"I forgot my password".  #1
#1
List of 10 synonyms for 'password'  #1
1. "Secret" 2. "Personal Identification Number" 3. "PIN" #2
4. "Access Code" 5. "Log-in Information" 6. "Log-in Data"  #2
7. "Account Information" 8. "Account Data"   #2
9. "Identification Number" 10. "Security Code"  #2

Inspect a complete higher-creativity synonym list showing both useful alternatives and outputs that need human judgment. section 7.2.1

How Effective Conversational AI covers it

How deep it goes
Defines temperature as a randomness and diversity control, but stops short of explaining its implementation or numerical scale.
Examples
Generates password synonyms, verb phrases, and transcript summaries while illustrating useful variation alongside oddities and hallucination risk.
What you can do after
Apply it via sampling-decoding settings to trade deterministic repetition for creative variation, then review outputs for relevance and hallucinations.
How it connects
Connects temperature with sampling decoding, greedy decoding, top-p, top-k, prompt engineering, and retrieval-augmented generation.

Also explained in

Compare how 17 books explain temperature →

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Effective Conversational AI. The section links point to the parts of the book it draws on.