CSCI 662 Spring 2026: NLP

🐝 Spring 2026 ⏰ Wed 2:00 - 5:20pm 📍 WPH 102

Announcements

The syllabus has been recently updated. Please familiarize yourself with the updates.

Catalogue Description

Natural language processing; language modeling; deep neural networks for language processing

Calendar + Syllabus

This calendar is subject to change. More details, e.g. lecture slides will be added as the semester continues. All work (except the project final report) is due on the specified date by 11:59 PM PT. See the syllabus for more details.

Course Description

This PhD-level seminar explores the research frontiers of modern Natural Language Processing, with a primary focus on large language models (LLMs). High profile technologies like ChatGPT brought NLP to the forefront of public discussion both inside and outside academia. But what underpins such technologies? The course is designed for PhD students prepared to engage with cutting-edge research literature and develop their own research ideas, regardless of prior formal NLP coursework. The course combines instructor-led lectures with student-led presentations and discussions of recent papers from top venues (e.g., ACL, EMNLP, ICLR, ICML, NeurIPS). Early lectures establish shared foundations for understanding LLMs, including model architectures, training and data paradigms, inference-time methods, and evaluation practices. Subsequent sessions focus on advanced and emerging topics such as reasoning and planning, adaptation and personalization, interpretability and controllability, robustness and generalization, data and evaluation methodology, and human–LLM interaction. The class will also explore details and variants of the real-world consequences of deploying large-scale LMs and NLP technologies more generally, such as the ethics and harms associated with them. Students are expected to independently read, analyze, and critique research papers, lead discussions, and complete a semester research project.

Learning Objectives

By the end of this course, students will be able to:

• O1: Apply key pieces of modern natural language processing pipelines, such as recurrent and Transformer-based sequence-to-sequence models.
• O2: Critically analyze contemporary research on large language models, including architectures, training and data paradigms, inference-time methods, and evaluation.
• O3: Explain and reason about the core principles, assumptions, and trade-offs underlying modern LLM-based NLP systems across topics.
• O4: Identify open research problems at the frontier of LLM research and lead scholarly discussions through presentation and contextualization of recent top-tier research papers.
• O5: Design, execute, and clearly communicate an original PhD-level research project on advanced NLP or large language models.

Assignments and Grading

There will be three components to course grades:

• Paper Presentation (20%).
  • Each student will complete one in-depth research paper presentation (15 mins presentation + 5 mins discussion) during the semester using slides, and will also serve as a designated discussion lead for two additional paper presentations. The presentation will be assessed on the student’s ability to clearly explain the paper’s motivation, technical approach, and key results; critically evaluate its assumptions, limitations, and contributions; and situate the work within the broader LLM and NLP research landscape (learning objective O2 & O3). Students will also be evaluated on the quality of the discussion they facilitate—through insightful questions, engagement with peers, and ability to surface open problems and future research directions—both during their own presentation and when leading discussions for others. The slides of the presentation need to be shared a day before the presentation to the class.
• Class Projects (65%).
  • Students will design and carry out a research project that aims to answer a question in natural language processing (Learning Objective O4). Students will work individually or in team of two or three. The focus of the class project can be research-focused or application-focused. A research-focused project will develop models and analyze data of an existing problem in NLP, or formulate a new problem altogether. An application-focused project will train (possibly only fine-tuning) and deploy NLP models to new application areas, while not necessarily developing any novel research question to be answered. Students will leverage tools, concepts, and techniques presented in the class. The project involves identifying a communication or exploration need that language could resolve, data sources available to inform the problem and method, and the techniques needed to approach it. The grading distribution for deliverables of the course project as well as expectations for each deliverable are detailed below. The deliverables include a project proposal (1 page single space) plus a project proposal pitch at the class, a mid project report (4 pages single space), final presentation (timed, with time for questions) and a final report (8 pages single space for the main document, up to 15 with appendix/figures). The final report will be due on the day of the University-scheduled final examination. Reports will be graded based on clarity, and completeness. The project is total 65% of final grade with the following breakdown:
  • Pitch + Proposal: 10%
  • Project Mid Presentation: 10%
  • Project Mid Report: 10%
  • Final Presentation: 15%
  • Project/Final Paper: 20%
• Class Participation (15%)
  • To encourage students to explain concepts underlying natural language processing in their own words (Learning Objective O2 & O3), we will evaluate each student’s engagements in course discussions during class and through the FAQ and research discussion during student’s paper presentations and project presentations. Points are earned by asking insightful questions during lecture and during project presentations, volunteering to present group work results during lecture activities, providing details and constructive answers to online questions on platforms like Piazza, and otherwise contributing to productive class discussions.

Grading inquiries and questions about the grading of the homeworks and the quizzes can be asked (to the instructor) within one week from the grading date (the date the grades are released). Grades will be available within 2-2.5 weeks after submission.

All written assignments related to the final project should use the standard *ACL paper submission template.

Grading Scale

• A: 94-100
• A-: 90-93
• B+: 87-89
• B: 83-86
• B-: 80-82
• C+: 77-79
• C: 73-76
• C-: 70-72
• D+: 67-69
• D: 63-66
• D-: 60-62
• F: 59 and below

Late Days

The course will employ a Late Day Token system that enables some flexibility with homework and project deliverables that are not in-class presentations. Note that, regardless of the use of Late Day Tokens, any assignment turned in 8 days late or more will receive an automatic zero. The course will allow for a budget of 5 Late Day Tokens per student. These tokens can be expended on homeworks, the paper review, and project deliverables (NOT quizzes or presentations or final project report) to extend the deadline, one day at a time, for a student without incurring a late penalty. These tokens should be used with no justification or explanation for taking the late time required (i.e., you do not need to explain your reason). Going over budget (e.g., turning things in late with no Late Day Tokens to expend) will incur grade penalties of 5% per day late. To ensure reasonable grading turnarounds and discussions of solutions, any assignment turned in 8 days late or more will receive an automatic zero regardless of the use of Late Day Tokens. For project teams, Late Day Token expenditures are on a per-student basis (i.e., if a team of 2 turns in their midterm report one day late, a member expending a Late Day Token will receive a 0% late penalty, while a member not expending a Late Day Token will receive a 5% late penalty). There are no refunds for late days: unused late days cannot be converted into credit of any kind.

Note: Please familiarize yourself with the academic policies and read the note about student well-being.

Pre-Requisites

Prerequisite(s): There are no formal prerequisites, however you should have a good understanding of the following (though we will review): Linear algebra (vector and matrix math basics); Probabilities: random variables, discrete and (some) continuous distributions, Bayes’ Theorem, Chain Rules; Calculus: mostly derivatives, or the ability to refresh this info; Programming: python using a Linux environment, mostly without a Jupyter notebook Co-Requisite(s): N/A Concurrent Enrollment: N/A Recommended Preparation:

• Fluency with Python programming.
• CSCI 567 (Machine Learning)
• CSCI 544 (Applied Natural Language Processing)

Course Evaluations

The course will follow the standard protocol for end of semester course evaluations. Time will be taken out of the final or penultimate week of class to enable students to complete these evaluations during class time under proctoring by a student volunteer. The instructor and TAs will not be present during the completion of these evaluations during class time, in accordance with University policy.

Similar Classes

• Graduate-level Applied NLP Fall 2024 CSCI 544
  • See highlighted class projects here.
• Undergraduate-level Special Topics: Language Models in NLP Spring 2024 CSCI 499
  • See previous class projects here.
• Undergraduate-level Special Topics: Language Models in NLP Fall 2023 CSCI 499
  • See previous class projects here.