You’re presenting your project in-class. Your final project write-up is due (on Gradescope) on a later date. See the schedule.

Guidelines for Final Write-up

You should submit:

  1. Your code: which is likely Python, but for some of you may be native application code (i.e., Swift)

  2. Any datasets and benchmarks that you build

  3. A write-up as a PDF. The write-up should include the following sections:

    • Introduction: What is the problem your project is trying to solve and a high-level overview of how you solve it and measure it. Give an example of a task that you can solve.

    • Methods: What is the technical approach that you take? This is where you should point to relevant portions of your code. You should include links, and also include small snippets of code when it makes sense to do so. If you’re using a complex prompt, you should include the prompt here. If you’re training a model, you should describe the training data and link to it.

    • Evaluation: Describe your benchmark and present results. Do not forget to mention the LLM that you use and the sampling parameters that you use for generation. You should have both quantitative and qualitative evaluation. The quantitative evaluation should summarize your benchmark results, e.g., with a mean pass rate. However, it is likely that the capabilities of your system go beyond what is easy to benchmark. Therefore, you should also have a qualitative evaluation, which has examples of both successful and failing interactions with your system. You may wish to include screenshots for the qualitative evaluation.

    • Discussion: This is where you may describe other approaches that you tried, or alternatives that you wish you had tried.

Guidelines for Presentation

  1. Before class begins, email me your slides (PowerPoint/Keynote/PDF)

  2. You will have 10 minutes to present your work, followed me 2-3 minutes for questions.

  3. Read the guidelines for the write-up. Your presentation guidelines are similar. I recommend allocating time as follows:

    • Introduction (2 min)
    • Methods (4 min)
    • Evaluation (4 min)