Announcement:
Learn more and sign up for the AgentX - AgentBeats Competition here.
Reminder: The Dec 8th class will start from 11:30am - 1:30pm PT.
Prospective Students
- Students interested in the course should first try enrolling in the course in CalCentral. Please join the waitlist if the class is full. The class number is 15131 for CS194-196 and 32761 for CS294-196.
- We plan to expand the class size to allow more students to join. Please fill in the petition form if you are on the waitlist or can’t get added to the waitlist. We expect to have accepted students enrolled 1-2 weeks into the Fall semester.
- Do not email course staff or TAs. Please use Edstem for any questions. For private matters, post a private question on Edstem and make sure it is visable to all teaching staff.
Course Staff
Teaching Staff: Xiuyu Li, Baifeng Shi, Chenyang Wang, Arhaan Aggarwal, Richik Pal
Guest Speakers
|
Yann Dubois Member of Technical Staff
|
Yangqing Jia VP, Al System Software
|
Jiantao Jiao Director of Research & Distinguished Scientist
|
|
Weizhu Chen Technical Fellow & CVP
|
Noam Brown Research Scientist
|
Sida Wang Research Scientist
|
|
James Zou Professor
|
Clay Bavor Co-Founder
|
Oriol Vinyals VP, Research
|
|
Peter Stone Chief Scientist at Sony Al, Professor at UT Austin
|
Class Time and Location
Lecture: 3-5pm PT Monday at Valley Life Sciences 2050
Course Description
Agentic AI is the new frontier and poised to transform the future of our daily life with the support of intelligent task automation and personalization. In this course, we will first discuss fundamental concepts that are essential for Agentic AI, including the foundation of LLMs, reasonsing, planning, agentic frameworks and infrastructure. We will also cover representative agent applications, including code generation, robotics, web automation, and scientific discovery. Meanwhile, we will discuss limitations and potential risks of current LLM agents, and share insights into directions for further improvement.
Syllabus
| Date | Lecture (3:10PM-5:00PM PT) |
Supplemental Readings |
|---|---|---|
| Sep 8 | Introduction Dawn Song, UC Berkeley [Slides] |
|
| Sep 15 | LLM Agents Overview Yann Dubois, OpenAI [Slides] [Recording] |
- KIMI K2: Open Agentic Intelligence - DeepSeek-V3 Technical Report |
| Sep 22 | Evolution of system designs from an AI engineer perspective Yangqing Jia, NVIDIA [Slides] [Recording] |
|
| Sep 29 | Post-Training Verifiable Agents Jiantao Jiao, NVIDIA [Slides] [Recording] |
- Introducing SWE-bench Verified - BrowseComp: a benchmark for browsing agents |
| Oct 6 | Agent Evaluation & Project Overview [Slides] [Recording] |
- Survey on Evaluation of LLM-based Agents |
| Oct 13 | Some Challenges and Lessons from Training Agentic Models Weizhu Chen, Microsoft [Slides] [Recording] |
|
| Oct 20 | Multi-Agent AI Noam Brown, OpenAI [Slides] [Recording] |
|
| Oct 27 | Predictable Noise in LLM Sida Wang, Meta [Slides] [Recording] |
- Adding Error Bars to Evals: A Statistical Approach to Language Model Evaluations |
| Nov 3 | AI Agents to Automate Scientific Discoveries James Zou, Stanford [Recording] |
- The Virtual Lab of AI agents designs new SARS-CoV-2 nanobodies - Paper2Agent: Reimagining Research Papers As Interactive and Reliable AI Agents |
| Nov 10 | Practical Lessons from Deploying Real-World AI Agents Clay Bavor, Sierra [Slides] [Recording] |
- τ2-Bench: Evaluating Conversational Agents in a Dual-Control Environment - Voice Sims: test agents in real world conditions before they talk to customers |
| Nov 17 | Multi-Agent Systems in the Era of LLMs Oriol Vinyals, Google DeepMind [Recording] |
|
| Nov 24 | No lecture — Thanksgiving week | |
| Dec 1 2:00-3:30pm PT |
Autonomous Agents: Embodiment, Interaction, and Learning Peter Stone, UT Austin / Sony AI [Recording] |
- Outracing Champion Gran Turismo Drivers with Deep Reinforcement Learning - SLAC: Simulation-Pretrained Latent Action Space for Whole-Body Real-World RL |
| Dec 8 11:30am-1:30pm PT |
Agentic AI Safety & Security Dawn Song, UC Berkeley [Livestream] |
Enrollment and Grading
Prerequisites: Students are strongly encouraged to have had experience and basic understanding of Machine Learning and Deep Learning before taking this class, e.g., have taken courses such as CS182, CS188, and CS189.
Please fill out the petition form if you are on the waitlist or can’t get added to the waitlist.
This is a variable-unit course. Detailed grading breakdown will be released soon!
Grading
| 1 unit | 2 units | 3/4 units | |
|---|---|---|---|
| Participation | 40% | 30% | 20% |
| Quizzes | 20% | 10% | 10% |
| Article | 40% | — | — |
| Project | — | Sum: 60% | Sum: 70% |
| Phase 1 | — | Sum: 45% | Sum: 50% |
| Proposal | — | 5% | 5% |
| Early demo (2-min video & milestone report) | — | 10% | 10% |
| Final Green Agent Submission | — | 30% | 35% |
| Phase 2 | — | Sum: 15% | Sum: 20% |
| White agent implementation | — | 10% | 15% |
| White agent report (1–2 pages) | — | 5% | 5% |
Bonus will be awarded for participating in AgentX-AgentBeats Competition.
Announcement for 1-Unit Students: You may choose to write an article or complete Phase 1 of the Agent Track. Either option is worth 40% of your grade. If you choose the Agent Track, coding requirements are minimal—building a simple agent (e.g., via prompt engineering) is sufficient.
Project Timeline
| Released | Due | |
|---|---|---|
| Project group formation | 9/15 | 9/22 |
| Phase 1 | ||
| Green agent proposal | 9/27 | 10/8 |
| Green agent demo submission & short report | 10/7 | 10/20 |
| Green agent submission — implementation, documentation & recording | 11/7 | 12/17 |
| Phase 2 | ||
| White agent final submission — implementation & report | 11/24 | 12/17 |
Article Timeline
Article (for 1-unit students) is due on 12/7.
Office Hours
-
Baifeng: Fridays, 5:30–6:30pm PT via Zoom(https://berkeley.zoom.us/j/5157669897)
-
Xiuyu: Thursdays, 8–9am PT via Zoom (https://berkeley.zoom.us/j/93276025385)
-
Richik: Time varies for each week. See Ed for OH announcements for each week. (https://berkeley.zoom.us/j/4342423164)