Paper Presentation & Discussion Questions (20%)

Each student will present one paper (typically with a partner) and lead an in-depth class discussion. Sign-ups will be available on BruinLearn and must be completed by the start of Week 2. Students who do not sign up by the deadline will be randomly assigned to a paper.

Each class session (90–110 minutes) is dedicated to a single paper. The two lead presenters are responsible for breaking the material down into 10–15 minute interactive reading and discussion segments. Presenters must submit their slides and discussion questions by 11:59 PM the day before their assigned presentation. Prior to each reading segment, presenters should provide specific prompts or tasks to the class to actively facilitate discussion.

Participation (20%)

Consistent attendance is required and will be tracked. Each student is permitted one unexcused absence without penalty.

Final Project Overview

The final project serves as the capstone of this course, challenging you to critically apply the concepts learned from reading papers to a concrete problem. Every project must successfully integrate a machine learning component with a robotics application, keeping the core debate of how we represent action (e.g., task-space vs. joint-space, continuous vs. discrete, hierarchical vs. end-to-end) at the forefront of your methodology.

We welcome a variety of project scopes to accommodate different research interests. You may choose one of the following tracks:

• Empirical & Experimental: Design, implement, and evaluate a novel policy or action representation on a simulated or physical robot.
• Theoretical: Formulate new mathematical frameworks, analyze algorithmic bounds, or investigate the theoretical properties of specific motion representations in robot learning.
• Comprehensive Survey: Write a thorough, publication-quality review paper on a targeted sub-topic (e.g., latent action spaces in foundation models). A successful survey must go beyond mere summarization to propose a novel taxonomy, synthesize trends, and identify critical gaps in the current literature.

Collaboration & Teams
Group work is highly encouraged. We recommend forming teams of 2–4 students. Collaborative projects naturally allow for a more ambitious scope, deeper technical execution, and richer discussions.

Building on Existing Research
You are welcome to build upon your ongoing lab research, provided it strongly aligns with the course themes. However, please note that pre-existing work will be evaluated with proportionately higher expectations. If you choose this route, your project proposal must clearly delineate what work has already been completed and what new, distinct contributions will be generated specifically for this class during the term.

Deliverables & Milestones:

• 1. Project Proposal (10%)
  Format: 1-2 pages. Submit via BruinLearn by 11:59 PM on the deadline.
  Your proposal should clearly articulate the problem or topic you are addressing and why it is relevant to the course. It must include a brief literature review, your proposed methodology, and a realistic weekly timeline. Crucially, tailor your methodology details to your chosen track: please specify your target environments/hardware (Empirical track), the mathematical frameworks you will analyze (Theoretical track), or your planned taxonomy and literature scope (Survey track).
• 2. In-Class Presentation (20%)
  Format: 12-minute presentation + 3-minute Q&A.
  During the final week of the quarter, your team will present your findings to the class. The presentation should clearly communicate your motivation, technical approach, and results. A crucial component of this presentation is a critical reflection on your chosen action representation—what worked, what failed, and why.
• 3. Final Report (30%)
  Format: 5–8 pages (excluding references), standard double-column conference format (e.g., IEEE, CoRL, or CVPR style). Submit via BruinLearn by 11:59 PM on the deadline.
  

  Your final report should be structured and written at the quality level of a formal academic conference submission. Regardless of your chosen track, the paper must critically evaluate the strengths, assumptions, and limitations of the specific action or motion representations central to your project.

  Expected Document Structure:

  • Abstract & Introduction: Clearly state the problem, your motivation, and your core contributions.
  • Related Work: Contextualize your project within the existing literature, explicitly drawing connections to the debates and papers discussed during the seminar.
  • Methodology: Provide a rigorous breakdown of your technical approach, mathematical framework, or the specific taxonomy used to organize your survey.
  • Evaluation & Results: Tailor this section to your track. Empirical projects must include detailed experimental setups, quantitative metrics, and qualitative analyses. Theoretical projects should present formal proofs, derivations, and algorithmic analyses. Survey projects should provide a deep synthesis of current trends and identify critical open challenges in the field.
  • Discussion & Conclusion: Critically reflect on your findings. What worked? What failed? How did your specific choice of representation fundamentally impact the outcome?

  Supplementary Materials: For empirical and theoretical projects, you are strongly encouraged to include a link to a GitHub repository containing clean, documented code and instructions for reproducibility.

  Reminder: All deadlines are firm. Under no circumstances will late final reports be accepted.

Letter Grades

Final letter grades will be assigned using the following scale, including plus and minus distinctions:

• 95% – 100%: A+
• 93% – 94.9%: A
• 90% – 92.9%: A-
• 87% – 89.9%: B+
• 84% – 86.9%: B
• 81% – 83.9%: B-
• 78% – 80.9%: C+
• 75% – 77.9%: C
• 72% – 74.9%: C-
• Below 72%: F