Advanced: Troubleshooting Variables

My variable is computing dates or time periods incorrectly

Brim includes an option for including today's date in the LLM prompt.

• If computing your variable requires knowledge of today's date, scroll down to the "Semantics" section, and make sure the box next to "Reference today's date" is checked.
• Explicitly specify the format in which you want a time variable and whether you want the variable to round up or down.

• Use the "timestamp" variable type for builtin formatting guardrails.

  
  

My variable is pulling a lot of irrelevant evidence

The LLM tries to find relevant evidence, and without more information, it can find too much.

• Try explaining the difference between what you want included and excluded. You can do this in the Variable Instructions, or by labelling examples in Label Review and Optimizing the variable.
• Be as black and white as you can in the instruction. Try adding the instruction to "Ignore" the specific case(s) you want to exclude.

• Limit generation by Document Type or Document Date if that is relevant for your variable. You can find these options in the "Semantics" section.

  
  

I'm not getting any results when I generate

By default, Brim hides label instances that represent an "Empty" response. In Label Review > Variables > choose a variable, try clicking "Show Default value labels".

If "Show Default value labels" is on and you see values now:

• Verify that your data set has evidence for this variable.
• Try making the variable name more descriptive.

If "Show Default value labels" is on and you still don't see values:

• Verify that your data set is loaded
• In the Variable Edit Screen, ensure that:
  • The "Include variable in generation" box is checked.
  • The "Restrict by Document Name" option is set to no Restrictions or correctly spelled Document names
  • The "Restrict by Document Date" option is set to no Restrictions or correct Document dates.
• Ensure that the patient you're looking for was included in your most recent label generation.

• Try generating again for this specific patient.

  
  

The LLM Reasoning for my variable is drawing the wrong conclusion

Vague or faulty LLM reasoning is often a consequence of the LLM not knowing how to behave in a specific situation:

• You can give in-line reasoning feedback and re-generate. Learn more here.
• Manually modify the prompt to add instructions for how to react if it is missing relevant information (e.g. "If there are no values for the input variable, return False.").
• Check that the variable has the correct evidence snippets.
  • If the evidence snippets are correct, focus on modifying the prompt to be more specific or clarifying the logic.
  • If the evidence snippets are incorrect, verify that important documents aren't being filtered out by title or date.

Running my variable is very slow or very expensive

The root cause of a very slow or expensive variable is usually the number of documents it's running on.

• Can you level-up the scope? Instead of one value per note, can your variable be one value per patient?
• Consider using conditional generation to limit the documents that the variable is run on.
• Can you run in batches?
• Do you have the "Use Advanced LLM Model" option checked in Variable Advanced Settings?