Lab Meeting Agendas

LM5, 9/30/2026 – start of RWeek 5

Announcement:

If possible, aim to produce some concrete results with a written description before the end of MP2. Remember you can use unlimited AI assistance for coding, experimentation, and writing.
- Don’t stress out about producing results, however. Grading will still primarily be based on effort.

Demos/presentations:

jmac demo 1: Getting AI help for planning an experiment
jmac demo 2: Using code completion via GitHub Copilot in VScode (essentially free and unlimited)
jmac demo 3: Agentic mode with GitHub Copilot in VScode
- You may need to pay for the $20 per month tier for this service
- The $20 per month Cursor offering may be better value – it’s easiest to work in their own development environment, not VScode.
John Lee: demo/discussion on “Tips and tricks for AI assistance in LLM research”
Jane Delassio chalk talk: “Bias in the ML Pipeline”

overview of current activities – random sample of about 5 people (I will paste the participant list into https://www.random.org/lists/)
Don’t forget the introductions channel
chalk talk(s) for next week???
Some nice results from Kevin:
- L2-SP, lambda sweep
- Can anyone else replicate or extend this? See the bullet point on L2-SP regularization at the end of the phonebook README.
end of MP4 is approaching – you may be disappointed with your grade if you do not log enough hours before Tuesday midnight to bring your average up to the minimum expectation.
chalk talk: Ty, “Comparing LLM performance for international humanitarian law”

introduction
a few highlights of the syllabus
WHP-TV CBS 21 news segment about DNU Lab, filming 9:00am tomorrow Thu Sept 3 at my office - all welcome - be a TV star!
“what should I do now?”
1. onboarding activities – scroll down to the README.
2. phonebook experiment – scroll down to the README.
3. Choose a follow-on experiment from the options listed at the end of the phonebook experiment README.
  - And another option is to read related literature.