### Answer --- LLMs don't always follow prompt instructions perfectly, even when they're clearly written. This is a known challenge with voice AI agents. [(1) ]() Here are likely reasons your agent is still saying "April 3rd 2026" instead of the numeric UK format: 1. Model Temperature Setting If your

model_temperature

is set higher (closer to 1) , the LLM behaves more randomly and may deviate from instructions. A lower value (closer to 0) makes responses more deterministic and instruction-following. [(2) ]() 2. Prompt Structure The Prompt Engineering Guide recommends using clear section headers and explicit formatting. [(1) ]() Try restructuring with stronger emphasis:

///## Date Format Rules (MANDATORY)
- ALWAYS use dd/mm/yyyy format
- NEVER say month names like "March" or "April"
- NEVER say ordinal dates like "3rd" or "4th"
- Example correct: "04/03/2026"
- Example incorrect: "April 3rd 2026"

3. Consider Conversation Flow For complex logic like strict date formatting, Conversation Flow agents can be more reliable because they handle tasks deterministically through node configuration rather than relying solely on LLM interpretation. 4. Use Dynamic Variables You could use a dynamic variable like

{{current_time}}

or

{{current_calendar}}

and pre-format the date before passing it to the agent, reducing the LLM's need to calculate or format dates itself. [(3) ]()

​ To learn more, see the sources I considered: [[1] Prompt Engineering Guide](https://docs.retellai.com/build/prompt-engineering-guide) [[2] Update Retell LLM](https://docs.retellai.com/api-references/update-retell-llm) [[3] Dynamic Variables](https://docs.retellai.com/build/dynamic-variables) ​

Feel free to tag @AI Evy with additional questions. I also appreciate feedback, it helps me improve.

### Answer Yes,

model_temperature

in the API is the same as "LLM Temperature" in the dashboard. [(1) ]() [(2) ]() The

model_temperature

parameter controls the randomness of the response, with values ranging from 0 to 1. [(1) ]() A lower value means more deterministic responses, while a higher value means more random. If unset, the default value of 0 applies. Since you have it set to 0.00, that's already the most deterministic setting available. [(2) ]() At this temperature, the agent should be highly consistent and follow instructions more reliably. If your agent is still not following the date format instructions despite temperature being at 0, the issue likely lies elsewhere—such as prompt structure, the specific model being used, or how the date calculation is being handled. You might try switching to a different model (like gpt-4.1) or using dynamic variables to pre-format dates before they reach the LLM. [(3) ]()

​ To learn more, see the sources I considered: [[1] Update Retell LLM](https://docs.retellai.com/api-references/update-retell-llm) [[2] Configure LLM Options](https://docs.retellai.com/build/llm-options) [[3] LLM Temperature Inaccuracy and Hallucinations](https://discord.com/channels/1186424376300679209/1404929409080754238) ​

Feel free to tag @AI Evy with additional questions. I also appreciate feedback, it helps me improve.