Stanford University · Course
Machine Learning from Human Preferences (CS329H)
Machine Learning from Human Preferences (CS329H) on Stanford Online is a ten-week graduate course on capturing human feedback to design the loss functions and rewards that are hard to specify directly, including reinforcement learning from human feedback for aligning language models.
Who it suits
- Researchers and engineers who work on alignment, language models or reward design.
- Learners who want first-principles foundations and want to implement state-of-the-art learning from human feedback.
What it covers
- Inverse reinforcement learning, which uses human preferences to specify a reward function.
- Metric elicitation, which uses human preferences to specify trade-offs for cost-sensitive classification.
- Reinforcement learning from human feedback, where preferences align a pre-trained language model.
Common questions about Machine Learning from Human Preferences (CS329H)
- How long is it?
- Stanford lists 10 weeks at 15 to 20 hours a week, for 3 academic units.
- What is the aim?
- Stanford says students should be able to understand and implement state-of-the-art learning from human feedback and be ready to do research on it.
- Can I enrol any time?
- Enrolment is only open during set periods, and an application is needed for graduate courses, so check the page for the current dates.
More like this
Keep learning on Infrato
- Path: AI for software developers
- More from Stanford University: all Stanford University listings in the directory
- Explore all AI courses
Where this comes from
Written from Stanford University’s own pages, read on 2026-10-14. Content, prices and availability change, so the provider’s page is the source. We are not affiliated with Stanford University and this is not an endorsement.