Lectures: MWF 2:00pm-2:50pm @📍PODEM 1A19
⚗️ Course Description
This course delves into the rapidly evolving field of large (foundation) models and machine reasoning. Large models, like large language models (GPT-4), video models (Sora), multi-modal models (GPT-4o, Gemini), world models, are creating impressive results in various domains. How do these models work? How can they be used to solve complex reasoning problems, in math, commonsense, embodied, and even scientific areas? Are we close to the so-called AGI already? What are the foundation limitations of these techniques? What are the next steps to build robust, grounded, and/or interpretable reasoning machines? We will talk about these topics. In addition to the lectures, students will also read and present latest related papers in the field, and team up to complete course projects.
Instructor: Lianhui Qin
Email: [email protected]
Office hours: Monday 3:30pm - 4:30pm
Office: CSE 2116
TA: Qiyue Gao
Email: [email protected]
Office hours: Tuesday 10am - 11am
Room: CSE B240A
TA: Jiannan Xiang
Email: [email protected]
Office hours: Thursday 3:30pm - 4:30pm
Room: CSE B240A
🚨 Announcement
<aside>
📢
</aside>
🗓️ Schedule / Syllabus (warning: subject to change 🌪️)
Schedule
🏆 Coursework ****
- Paper Presentation: Starting from week 3, each student will give a presentation on a paper related to large model reasoning. The presentation will be 5 minutes long.
- Course Project: There's a course project that you'll be working on throughout the quarter. You are strongly encouraged to form teams of 3-4 students for the final project. Larger groups are expected to deliver larger projects. All reports should use NeurIPS template.
- Late Days: You will have 2 penalty-free late days in total for the course project (max 1 late days per report). Any delayed submission after the late days will be penalized 10% per day. Late days will be applied to all members in the project teams. Note that the late days DO NOT apply to final report.
- Deadlines: All deadlines are 11:59pm PT.
- Submission & Grading: Please submit your assignments to Gradescope. If you haven't signed up an account yet, please use your @ucsd.edu email to sign up. Grades will be released via Gradescope too.
Paper Presentation (26%)
- Please sign up for the presentation in the given google spreadsheet (1%).
- You may choose a paper from the reference list or select another paper relevant to the course topic. If you wish to present a paper that is not included in the reference list, please add it to the "Propose Your Own!" tab in the spreadsheet for TA approval.
- [Instruction]
Course Project (70%)
- Identify team members & project proposal (7%): Oct 25 (Week 4 Friday)