Zero shot classification

[abratnap]

Describe the contribution to the taxonomy

• Adds a skill for classifying user queries based on class descriptions
• Zero shot classification which can be tuned based on description

Input given at the prompt

   ...

Response from the original model

  ...

Response from the fine-tuned model

  ...

Contribution checklist

• The contribution was tested with lab generate
• No errors or warnings were produced by lab generate
• All commits are signed off (DCO)
• The qna.yaml file contains at least 5 seed_examples
• The qna.yaml file was linted and prettified (yaml-validator can do both)

[mingxzhao]

We have changed the attribution requirements and no longer require them within the yaml file. Please remove them when you can!

[jjasghar]

This feels like it should be a knowledge with references instead of context on it. The context is written and opinionated, where a knowledge would be referenced trusted information.

[abratnap]

This feels like it should be a knowledge with references instead of context on it. The context is written and opinionated, where a knowledge would be referenced trusted information.

The concept is to enhance the ability to classify text into categories based on class descriptions in real-time. While LLM can already manage this, I believe adding a specific skill would further improve its performance in such tasks.

[jjasghar]

We will be discussing this in our skill triage meeting soon. Thank you for the feedback.

Can you rebase the branch too please?

[jjasghar]

We have decided that the "guard rails" for this type of work should come from the core engineering team. Thank you for your submission, but this will PR will fall into a larger initiative, and we can not accept this.

[abratnap]

We have decided that the "guard rails" for this type of work should come from the core engineering team. Thank you for your submission, but this will PR will fall into a larger initiative, and we can not accept this.

okay. thank you for reviewing it.