Students are required to conduct and submit a research project as well as complete a project presentation in an assigned lecture slot during Week 9. Students may work on projects in groups of up to 3 people. Note the topic of your project can be different from the topic of your presentation.
The project can be on any topic related to the course material. The purpose of the project is to gain a better understanding of experimental and/or theoretical aspects of active representation learning algorithms. While we encourage you to explore new research directions, projects may also focus on reproducing or comparing previous approaches, perhaps investigating various theoretical or design choices. If you are a graduate student currently working on a related topic, you may use your research for the project. Generally, projects will fall into one of the following categories:
Theory: study proof techniques, try to extend a proof, or apply a proof to a new setting.
Algorithms & Models: extend an algorithm or model, design a new one, or adapt one for a new setting.
Applications: identify the correct assumptions for approaching a new setting, experimentally validate these assumptions.
Due April 20th.
The project proposal is a mechanism to get you to start thinking about your project. It is non-binding; you can change the direction of the project after the proposal submission.
Project proposals could be 2-3 pages (3 pages maximum), excluding references.
The structure for the project proposal should generally cover the following aspects:
Introduction of research area: Motivate why this area is interesting and describe key technical challenges.
Related work: Give an overview of different types of approaches and highlight assumptions of previous work. This should not be a laundry list of previous papers. Rather, it should describe previous approaches in broad strokes.
Research question(s): Describe one or more research questions to be addressed. The research question should be a high-level description of the project topic.
Research plan: Describe a preliminary plan to address the research question(s), and outline any experimental set-ups, such as datasets, models, etc. The research plan should extend the research question(s) to the details of the project.
Proposal submission will be done via Gradescope.
Your project may focus on any topic related to the course content on active representation learning, and you are encouraged to integrate the project closely with your own research interests. Potential project directions include, but are not limited to, the following:
Scalable Deep Active Learning: Design and implement scalable acquisition strategies for deep active learning in real-world scenarios with limited labeled data. Identify domain-specific challenges such as labeling costs, data complexity, or computational constraints. Evaluate how leveraging these domain insights improves performance compared to traditional supervised methods. Relevant application domains include medical imaging, computer vision tasks, or active perception in robotics.
Generative Active Learning: Explore approaches for synthesizing data points to optimize active learning efficiency. Investigate how generative models can augment or replace real data collection, especially when labeled data is costly or difficult to obtain. Evaluate the effectiveness and efficiency gains of your methods on realistic datasets or simulations, such as robotic manipulation tasks in simulation or augmenting text datasets in NLP.
Representation Learning for Bandits: Develop novel algorithms for exploration in high-dimensional spaces using neural bandit approaches. Analyze and empirically evaluate how effective representation learning can facilitate better exploration strategies compared to conventional bandit methods. Potential applications include recommender systems, personalized content selection, or structured scientific experimental design.
Goal-oriented Exploration and Representation Learning: Implement and evaluate algorithms for active multi-task representation learning, goal-conditioned curiosity-driven exploration, or curriculum learning. Your project might involve devising strategies for goal generation, investigating task relevance, or studying the role of intrinsic motivation in complex decision-making scenarios.
Active Fine-tuning of Foundation Models: Develop methods for actively selecting examples or user feedback for efficient fine-tuning of large foundation models such as large language models (LLMs), vision transformers, or multimodal models. Concretely, explore techniques for active demonstration selection in few-shot NLP tasks, active preference elicitation for aligning chatbot responses, or active data sampling for fine-tuning vision models on downstream tasks. Evaluate improvements in model performance, data efficiency, and practical alignment outcomes in relevant benchmarks or user studies.
The final project report should resemble a (mini-) research paper in scope and tone to the papers discussed in class. You are not expected to produce state-of-the-art results or exhaustively explore every aspect of the problem. However, your report should reflect thoughtful engagement with the topic — in how you motivated the problem, designed your approach, and evaluated your results.
No strict page limit. Most reports are approximately 5–8 pages, with additional pages permitted for references, appendices, figures, etc. We do not have strict formatting requirement. You may use the NeurIPS LaTeX format (with the "preprint" option).
There is no formal grading rubric. Your report will be evaluated based on the depth of thought in three areas: problem setting, approach, and investigation. We will look for a clear motivation of an interesting or challenging problem, a thoughtful approach that engages with prior work and addresses its limitations, and a careful reflection on any implementation challenges. Additionally, we expect meaningful evaluation metrics, comparisons to relevant baselines, and insightful analysis or visualization of results. For theoretical projects, you should also clearly state assumptions and limitations.
You may organize your report however you prefer. A recommended outline is as follows:
Introduction
What is the motivation for this project?
What are the key technical challenges?
Briefly summarize your contributions.
Include GitHub links, website, or videos, if available.
Background and Related Work
Review essential background.
Highlight previous approaches and their assumptions/limitations.
Avoid simply listing papers — synthesize insights.
Problem Statement
What questions are you exploring?
Methodology
Justify your chosen methodology, and include derivations or proofs here (or defer details to an appendix).
Experiments and Results
Describe the dataset(s), models, training setup, and evaluation metrics.
Present your findings using figures/tables. Include additional experimental details in appendices if needed.
Discussion and Conclusion
Recap the research question, your approach, and main findings.
Reflect on what worked, what didn’t, and why.
Suggest directions and timeline (if applicable) for future work.