To get started with your project, we’d like you to think about some of the ways people are currently using language models or could be using them in the near future if they worked better.
You should pick one of these current or prospective use cases of language models to delve into over the course of the semester.
Your chosen use case should go beyond the applications that are covered in class.
By this, we mean you should select a use case that does not appear to be covered in any of the lectures, OR if you draw from application that is being covered, you should plan to delve into your chosen use case in substantially more depth than you expect us to in class.

We use the term “use case” rather than “application” because we’d like you to hone in one specific LLM capability for your project. Here are a few examples to give you a flavor of the types of use cases we think could make good course projects.

  • If you are interested in LLMs + code, you might choose to do your project on:
    • Code autocomplete
    • Vibe coding of websites
  • If you are interested in LLMs + agents, you might choose to do your project on:
    • Websearch agents that browse the internet on your behalf
    • Deep research agents that aim to assist with academic literature reviews
  • If you are interested in LLMs + creativity, you might choose to do your project on:
    • Co-creative writing or storytelling through chain writing
    • A tool for brainstorming, ideation, or concept generation (e.g., for plots, games, or product ideas)

You may choose between two tracks for the project:

  1. Research track: Pose a novel research question about the efficacy of different methods for your chosen use case. By the end of the semester, you should have designed and run a set of experiments that either support or reject your initial hypotheses.
  2. Applied track: Build a demo of an LLM system for your use case. By the end of the semester, you should have a working demo that the instructors can interact with. You should be prepared to explain each of the LLM system components that make up your demo.

The project may be completed either individually or in a team of two students. Teams of two will be expected to have more ambitious project plans than teams of one.

Initial Pitch (Due 11:59 PM on 1/23)

Please submit on Canvas the following:

  1. Names and AndrewIDs of all group members.
  2. What is the title of your proposed project? (keep these <~10 words please)
  3. In a few sentences, describe the use case you have selected. What does user interaction look like–that is, what do users input to the system and what do they expect as output?
  4. Are you planning to pursue the research track of applied track?
  5. List a couple academic papers that describe how to use LLMs for this use case.
  6. If applicable, list any non-academic efforts you found that have made efforts on this use case (e.g. startups, big tech efforts, anonymous social media posts, etc.).

Our goal with the initial pitches is to be able to give you early feedback on whether your idea is well-scoped and likely to lead to a successful project.

Proposal (Due 11:59 PM on 2/12)

Please submit on Canvas the following:

  1. Literature review: Conduct a literature review of academic papers that attempt to tackle the use case you have described. What techniques do they use? How do they evaluate model performance?
  2. Industry review: Review any non-academic efforts trying to tackle this use case. What do these systems look like? Describe any details that are available about how they work.
  3. Proposed method: Describe in detail your proposed method, datasets, and evaluation scheme.
  4. Midpoint goals: What does success at the midpoint check-in look like for your project? Describe both key progress you expect to make by then, as well as some of your reach goals.
  5. Overall goals: What does success for your project look like by the project end? Describe both key progress you expect to make in order to achieve full marks, and reach goals that you are excited by but may be beyond the scope of a 3-month project.
  6. Collaboration plan: Only fill this out if you are working in a team of two. How do you plan to divide up the labor for the project? Who is responsible for what?

Midpoint check-in (Due 11:59 PM on 3/12)

TBD

Final Report and Presentation (Due before class on 4/21)

TBD