Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Week 4 presentation proposal #2397

Merged
merged 1 commit into from
Sep 11, 2024

Conversation

leegrash
Copy link
Contributor

@leegrash leegrash commented Sep 2, 2024

Assignment Proposal

Title

A/B testing of LLMs to increase user experience on a platform using Optimizely

Names and KTH ID

Deadline

  • Week 4

Category

  • Presentation

Description

We want to show the advantages of refining the user experience of LLMs on a platform by A/B testing using the tool Optimizely. The presentation will show the strengths of using A/B testing to analyze usage of the platform and show metrics that can be gathered through Optimizely to tune the LLM specifically to the platforms needs.

Relevance

We think that understanding comparative testing methods is important to being able to actually use ML tools in practice. According to the MLOps wikipedia page, very few users of MLOps-practices actually found any real world benefits and being able to reliably test is key in determining whether or not to use it.

@sofiabobadilla
Copy link
Collaborator

Hi @leegrash
The idea of the proposal has potential, yet I would like you to improve both the description and the proposal focusing on how this will be connected to MLOps.

As it is now, the description makes me think you are working on the wrong week (week 2), while it should be week 4.

FYI: We have seen A/B testing today on week 2 so consider this when designing the presentation.

@leegrash
Copy link
Contributor Author

leegrash commented Sep 8, 2024

Hi!

I understand. The way we see it, A/B testing could either be used to test different iterations LLMS on a platform with users. This is one way we could focus our presentation. Another way is to focus on the systematical testing advantages of LLMS with incremental changes to variables. What way would you suggest us to focus in order for us not to overlap with other presentations or topic weeks?

@sofiabobadilla
Copy link
Collaborator

A/B testing for LLMs by itself is not very original.
But if you talk about A/B testing for LLMs using X tool that can be more interesting.

For example this talk from week 2: Property-based testing in Python using Hypothesis

Also, in 2020 someone presented a similar topic (link) . Maybe you can get more inspiration for the presentation from it :)

@leegrash
Copy link
Contributor Author

leegrash commented Sep 9, 2024

I understand. I updated the topic proposal to be more specific regarding tools and purpose. Is this okay?

@sofiabobadilla
Copy link
Collaborator

The proposal is clear now.
I'll merge.
@leegrash, FYI: I'll send you a calendar invite for your case
Good luck!

@sofiabobadilla sofiabobadilla merged commit a5d5045 into KTH:2024 Sep 11, 2024
2 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants