logistics

Textbooks

Course readings will primarily include published research papers, provided as PDFs.

Additional useful deep learning and LLMs texts include:

Grading criteria

The tentative grading breakdown is as follows:

  • Quizzes (10%)
  • Homework assignments (50%)
  • Final project (40%)

Late policy

Each student is allowed two late days total across all homework submissions (not per homework). Once these two days are used, no additional late submissions will be accepted. If you encounter an unexpected health or personal emergency, please contact the instructors as soon as possible at cs5624instructors@gmail.com, and provide valid documentation if needed. Late submissions due to job interviews, other coursework, or non-emergency situations will not be excused, and extensions will not be granted.

AI assistance policy

AI assistance is permitted for completing assignments, including quizzes, homework, and projects. However, if you use AI tools, you must submit the prompts you used and describe how the AI contributed to your work. It is your responsibility to verify the AI-generated content for accuracy before submission.

Student collaboration policy

For group presentations, work must be completed solely by members of the group. Apart from group presentations, students must independently complete their work. Sharing code or written materials with others is prohibited. However, discussing concepts, clarifying instructions, exploring problem-solving approaches, or seeking debugging advice (without sharing code) is encouraged. All external sources must be properly cited.

Academic accommodations

Virginia Tech is committed to creating an inclusive learning environment where all students have equitable access to educational opportunities. If you anticipate or experience barriers in this course due to a disability—including, but not limited to, chronic health conditions, mental health conditions, learning disabilities, mobility disabilities, sensory disabilities, ADHD, or other disabilities—I encourage you to connect with Services for Students with Disabilities (SSD).

SSD works collaboratively with students to determine reasonable accommodations that provide equal access while maintaining the essential learning outcomes and academic standards of the course.

If you have not yet connected with SSD, you may contact their office at 540-231-3788, ssd@vt.edu, or visit www.ssd.vt.edu for more information about the accommodation process.

If you have received an SSD accommodation letter, please meet with me privately as early in the semester as possible so we can discuss how your accommodations will be implemented in this course. Accommodations are not retroactive, and reasonable advance notice is needed to coordinate them. In most cases, this means providing at least three (3) business days’ notice for routine accommodations and ten (10) business days’ notice for final exams.

I value creating a classroom environment where all students can fully participate and succeed, and I welcome conversations about how to support your access to this course.

Graduate Academic Integrity Statement

The tenets of the Virginia Tech Graduate Honor Code will be strictly enforced in this course, and all assignments shall be subject to the stipulations of the Graduate Honor Code. For more information on the Graduate Honor Code, please refer to the GHS Constitution.

Course/learning objectives

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

  • Explain foundational NLP concepts, key architectures underlying modern LLMs, including neural language models, Transformers, and multimodal and diffusion-based models, and the design and operation of LLM agents.
  • Analyze recent advances in LLM pretraining and post-training, efficient training and inference, reasoning and test-time scaling, interpretability, evaluation, security, and privacy.
  • Evaluate emerging approaches for AI agents and self-improving systems, including recursive self-improvement, harness evolution, continual learning, and automated scientific discovery.
  • Critically read and assess contemporary research papers, identifying assumptions, limitations, and open problems.
  • Formulate well-defined research questions, and develop and evaluate principled solution approaches through implementation and experimentation.

Absence policy

Although attendance is not graded, students are strongly encouraged to attend all scheduled classes and project presentations to fully engage with the course material and succeed.

VT principles of community statement

Virginia Tech is a public land-grant university, committed to teaching and learning, research, and outreach to the Commonwealth of Virginia, the nation, and the world community. Learning from the experiences that shape Virginia Tech as an institution, we acknowledge those aspects of our legacy that reflected bias and exclusion. Therefore, we adopt and practice the following principles as fundamental to our on-going efforts to increase access and inclusion and to create a community that nurtures learning and growth for all of its members:

  • We affirm the inherent dignity and value of every person and strive to maintain a climate for work and learning based on mutual respect and understanding.
  • We affirm the right of each person to express thoughts and opinions freely. We encourage open expression within a climate of civility, sensitivity, and mutual respect.
  • We affirm the value of human diversity because it enriches our lives and the University. We acknowledge and respect our differences while affirming our common humanity.
  • We reject all forms of prejudice and discrimination, including those based on age, color, disability, gender, gender identity, gender expression, national origin, political affiliation, race, religion, sexual orientation, and veteran status. We take individual and collective responsibility for helping to eliminate bias and discrimination and for increasing our own understanding of these issues through education, training, and interaction with others.
  • We pledge our collective commitment to these principles in the spirit of the Virginia Tech motto of Ut Prosim (That I May Serve).