Course Syllabus

This course focuses on solving natural language processing (NLP) problems using large language models (LLMs). Our goal is not to showcase how to apply LLMs as off-the-shelf tools, but to explain in detail how they are developed so that students can better understand their behavior and use them more effectively for various tasks. Core topics include transformer architectures, transfer learning, instruction following, alignment methods, and evaluation. The course also covers efficiency considerations such as quantization, parameter-efficient finetuning, and KV cache. Applications explored include text classification, text generation, machine translation, summarization, information retrieval, and question answering. This course does not cover the classical NLP pipeline and linguistic structure prediction, computational social science, or how to use computational methods to improve the scientific understanding of natural language.

This course is designed for computer science/computing majors. We assume prior experience in machine learning and proficiency in Python and PyTorch. Students should be familiar with neural networks, PyTorch, and NumPy; no introductory tutorials will be provided. In previous offerings of CS 6340/5340, we revisited this background over the course of two weeks. Since we have introduced CS 3350 or CS 5353 as prerequisites for the undergraduate section, these lectures are dropped in Fall 2026.

We expect that you are comfortable with basic calculus, probability, and linear algebra. You should be comfortable taking (multivariable) derivatives and understanding matrix/vector notation and operations. You should know the basics of probability, mean, standard deviation, etc.

Revisiting/polishing your knowledge. You can prepare by:

1. There are a ton of Python resources for people with some programming experience. Check them out hereLinks to an external site.. I highly recommend these slidesLinks to an external site./colabLinks to an external site. (caution: there are some typos) and my colleagues suggest these: 1Links to an external site., 2Links to an external site., 3Links to an external site., and 4Links to an external site.. 

2. Math and machine learning basics are nicely covered in the first part of the Deep Learning bookLinks to an external site.. Obviously, you can use the same book to familiarize yourself with deep learning. If you learn by coding then you will find this resource helpful Practical Deep Learning for Coders by Fast.aiLinks to an external site..

The following texts are useful (especially the first one), but none are required. All of them can be read free online.

• Dan Jurafsky and James H. Martin. Speech and Language Processing (3rd ed. draft)Links to an external site.
• Jacob Eisenstein. Natural Language ProcessingLinks to an external site.
• Yoav Goldberg. A Primer on Neural Network Models for Natural Language ProcessingLinks to an external site.
• Ian Goodfellow, Yoshua Bengio, and Aaron Courville. Deep LearningLinks to an external site.
  Links to an external sitLinks to an external site.

On the class websiteLinks to an external site., readings are assigned to complement the material discussed in each lecture. You may find it useful to do these readings before the lecture as preparation or after the lecture to review, but you are not expected to know only what is covered in the lecture! 
Lectures will be recorded and shared sometime after they take place. However, this is not an online course: office hours might not be offered remotely, evaluation will be entirely in person, recording may fail due to technical difficulties and no alternatives will be provided in such cases, uploading recordings may be delayed, parts presented on the whiteboard may not be recorded, etc. 

Join Piazza: https://piazza.com/utah/fall2026/cs63405340fall2026

Note:

• Announcements about scheduling, released grades, and other logistical issues will be made through Piazza.  
• We are open to appointments outside of our office hours, but make sure to reach out a few days in advance so that we can find a time that works for someone on the course staff. 
• The instructor and the TA will aim to respond within two business days.
• The instructor and the TA may not read email or check Piazza after ~5–6pm.
• The instructor does not read email or check Piazza on weekends or on holidays, and encourages TAs to do the same. 
• You can post anonymously to other students on Piazza, but your name will always be visible to the instructor and the TA.
• Feel free to answer questions!

Zoom Etiquette

• Come prepared with your questions. Just as you would in person, have your topics or concerns ready before joining.

• Avoid joining and leaving repeatedly. Avoid joining, asking one question, turning off your camera, and waiting silently; or joining and leaving repeatedly. In a physical office, you wouldn’t linger until another question came to mind – treat Zoom the same way.

• Turn on your camera when possible. It helps create a more natural, interactive experience.

Help with programming tasks during office hours:

• TAs will follow my guidelines, which emphasize not looking at and debugging the code directly, but having students go over their approach and lead them in the right direction through a conversation. Otherwise, this quickly turns into giving away solutions, which is unfair, doesn’t promote a deeper understanding, and doesn’t enhance problem-solving skills. Machine learning debugging is notoriously challenging. To avoid difficulties, start your programming tasks early.

• Please arrive on time. If you're late, enter quietly. If you miss part of a lecture or announcement, don't ask me to repeat it. Please check with a classmate about anything important you missed; or in the recording.
• No working on laptops. Ideally, this would be a device-free classroom, but note-taking on a tablet is fine.
• No large meals in the classroom; snacks are okay.

• Remote office hours / Zoom:

  • Come prepared with your questions. Just as you would in person, have your topics or concerns ready before joining.

  • Avoid joining and leaving repeatedly. Avoid joining, asking one question, turning off your camera, and waiting silently; or joining and leaving repeatedly. In a physical office, you wouldn’t linger until another question came to mind – treat Zoom the same way.

  • Turn on your camera when possible. It helps create a more natural, interactive experience.

Your performance in this course will be evaluated by:

• Programming (in person): 50% of your overall score, based on the average of your percentage scores. Programming will be organized as follows:

  • Practice assignments will be given in advance; however, they will not be graded for points. Their purpose is to give you practice, expose you to the problem, and prepare you for the in-class implementation.

  • During class, you will complete a related programming task, using your own laptop in a provided programming interface with AI-assisted autocomplete disabled. The in-class implementation will be graded and will constitute the programming portion of your course grade. The in-class tasks will be similar to the take-home assignments but smaller in scope and may differ in details. 

  • If you don't have a working laptop, you can request it from the library: https://lib.utah.edu/services/knowledge-commons/checkout-equipment-faq.phpLinks to an external site. 

• Quizzes (in person, on paper): 50% of your overall score. Quiz grades are based on total points earned out of total points possible, not the average of individual quiz percentages. Quizzes may differ in point value, so longer quizzes count proportionally more toward the quiz grade. 

There is no midterm or final exam.

Attendance in person is mandatory for quizzes and programming evaluation, but not for lectures. Quizzes and programming sessions may not be taken at a different time for any reason other than a medical emergency or conflict with another exam, and documentation may be required. Do not make non-academic travel plans until after the final exam period has ended.

My plan for quiz/practicum schedule is here: https://drive.google.com/file/d/1GSPm57Dt8FT9kTI4O8etrFgnrNHy-4qT/view?usp=sharing (use @gcloud.utah.edu account)

Long Accommodations 

Accommodations will be considered on an individual basis and may require documentation. Please contact your instructor as soon as possible (preferably shortly before the semester begins) to request accommodations of any kind.

Accommodations for In-Person Graded Components

Attendance in person is mandatory for quizzes and programming evaluation, but not for lectures. Quizzes and programming sessions may not be taken at a different time for any reason other than a medical emergency, conflict with another exam, or presentation at a major academic conference.  Documentation may be required. Do not make non-academic travel plans until after the final exam period has ended.

General Accommodations for Undergraduate Students

There will be some, but details still TBD

Extreme personal circumstances

Please contact your instructor as soon as possible if an extreme personal circumstance
(hospitalization, death of a close relative, natural disaster, etc.) is interfering with your ability to
complete your work.

Religious Practice

To request an accommodation for religious practices, contact your instructor at the beginning of the semester.

Active Duty Military

If you are a student on active duty with the military and experience issues that prevent you from participating in the course because of deployment or service responsibilities, contact your instructor as soon as possible to discuss appropriate accommodations.

Disability Access

All written information in this course can be made available in an alternative format with prior notification to the Center for Disability Services (CDS). CDS will work with you and the instructor to make arrangements for accommodations. Prior notice is appreciated. To read the full accommodations policy for the University of Utah, please see Section Q of the Instruction & Evaluation regulations.

If you will need accommodations in this class, contact:

Center for Disability Services
  801-581-5020
  disability.utah.edu
  162 Union Building
    200 S. Central Campus Dr.
     Salt Lake City, UT 84112

If you believe there is an error in grading, you may request a regrading within one week of receiving your grade. Requests must be made through Gradescope, explaining clearly why you think your solution is correct. Please maintain a professional and respectful tone – the course staff can make mistakes, and we will fix mistakes.

No additional chances for earning extra credit will be granted. If you're finding it difficult to keep up, please don't hesitate to reach out to us or graduate/undergraduate/CDA advisors. We're here to help you find ways to better manage upcoming deadlines.

We plan to map numerical grades to letter grades at the standard scale (see below) and do NOT plan to round or curve the grades.

University of Utah grading scale:

The class operates under the Kahlert School of Computing’s  and the John and Marcia Price College of Engineering's policies and guidelines, including the academic misconduct policy:

• College policies: https://www.coe.utah.edu/students/current/semester-guidelines/
• School policies: https://handbook.cs.utah.edu/2024-2025/CS/Academics/policies.php  
• Academic misconduct policy: https://www.cs.utah.edu/undergraduate/current-students/policy-statement-on-academic-misconduct/ 

More information about the academic misconduct policy can be found on the linked page, but we highlight this: 

Violations of this policy are recorded as “strikes” by the SoC. A failing grade sanction in an SoC course counts as one “strike” in a student’s academic record. Two lesser sanctions in SoC courses count as one “strike” in a student’s academic record. Any student with two strikes due to academic misconduct will be subsequently barred from registering for any additional SoC courses, immediately dropped from their respective degree program, and will not be admitted to any future SoC program.

As defined in the University Code of Student Rights and Responsibilities, academic misconduct includes, but is not limited to, cheating, misrepresenting one’s work, inappropriately collaborating, plagiarism, and fabrication or falsification of information. It also includes facilitating academic misconduct by intentionally helping or attempting to help another student to commit an act of academic misconduct. A primary example of academic misconduct would be submitting as one’s own, work that is copied from an outside source.

Academic misconduct in this course will result in a failing grade for the specific quiz or programming activity involved. 

We don't forbid using generative AI tools for ungraded take-home assignments, but you can't use them during in-class programming evaluations. So be cautious about overreliance – don't mistake productive code generation for understanding, or you won't be prepared for the in-person programming.

I recommend checking this out: https://cte.utah.edu/instructor-education/ai-website-content/student-ai-guide-2025.pdf 

We recognize that students experience personal and/or academic challenges during the semester. Your health and well-being matter, and the University of Utah offers many resources to support you, including counseling, wellness programs, crisis services, and more. You can explore these at: https://wellness.utah.edu/for-students.php.

Please note that while we care about your well-being and are happy to help connect you with appropriate resources, we are not therapists or counselors. Some types of personal disclosure, especially involving traumatic experiences, can be emotionally difficult for us to receive and are best shared with trained professionals who can provide the care and support you deserve. If you are in crisis or need someone to talk to, we strongly encourage you to reach out to the campus resources above or another mental health professional.

This syllabus is not a contract. It is meant to serve as an outline and guide for your course. Please note that your instructor may modify it to accommodate the needs of your class.

You will be notified of any changes to the Syllabus.