Issue details: We are using Gemini Flash 002 for doing image and video analysis and predict content details. The output if in form of tags that we are comparing against a fixed set of vocabulary, given by our business.
Ex: Image of a table with food items.
LLM Output: { Items: "Apple ", "Beverage", "Soups" }
The issue we are facing is that sometimes LLM is giving SINGULAR form a an Item even if more than one exist. Example: Its giving "soup" instead of "soups" , "beverage" instead of "beverages"
We need a way to makes sure always PLURAL forms are returned even when their are singular quantities available. Have tried many changes to prompt but no luck so far.
Let me know if anyone has faced similar issue and possible approaches to fix this.
LinkedIn
Twitter
