Skip to content
Scalable Trustworthy AI Scalable Trustworthy AI

STAI Lab - KAIST AI

Build your research career at STAI.

Research AI safety, agents, memory and retrieval at KAIST AI, with direct mentorship from Seong Joon Oh.

Apply to STAI Explore openings See our publications

Research focus

AI safety, agents, memory and retrieval.

Build useful AI that people can guide, correct and trust.

Knowledge-Intelligence Separation asks what belongs in model parameters, and what belongs in editable memory and tools.

Agents and system evaluation

Improve the whole system, from planning and tool use to coordination between agents.

Questions and papers on Agents and system evaluation
  • When an agent fails, was the cause its plan, a tool, retrieved evidence or communication with another agent?
  • How can we compare systems and check their reasoning and actions with reproducible evaluations?

Explore our work

Memory and continual adaptation

Help agents learn from new information and user corrections without retraining for every update.

Questions and papers on Memory and continual adaptation
  • How should an agent update or delete a memory and revise the facts that depend on it?
  • How can it recognise that a previously known answer has become unknown, rather than repeat an outdated fact?

Explore our work

What would you like to investigate? Tell us in your application.

Apply to STAI Explore all publications Read our research vision

Publication record

Research with a publication record.

17 conference papers in 2026.

4 NeurIPS 2026 1 EMNLP 2026 5 ICML 2026 2 ACL 2026 5 ICLR 2026

Conference totals include main, Findings, demo and dataset tracks. Workshops and preprints are excluded.

Explore the full publication record

Student research

Your first paper can start here.

Recent first-author papers by our MSc researchers and alumni.

Undergraduate? Start with an internship.

Work with a PhD student, develop research skills and aim to contribute to a paper as a co-author.

Apply for an internship How internships work Meet potential mentors

Mentorship

Direct support. Independent research.

1:1 time with Seong Joon Oh

Discuss your ideas, results and next steps directly with your supervisor.

Additional meetings

Book extra time with Seong Joon Oh when you need feedback.

Lead your own project

Graduate students lead first-author projects. Interns work closely with a PhD student host.

Our research principles

Read our expectations, PhD milestones and approach to researcher development.

Read our principles

Alumni

Move forward. Stay connected.

Our alumni build careers in industry and academia, and continue to publish with us.

Build international collaborations.

Connect through alumni projects, external co-supervision, research internships and conferences.

Meet our alumni Collaborate with us
STAI lab members on a hike at Cheonggyesan Mountain
Away from our desks. Cheonggyesan, May 2026.

Life at STAI

Meet your future colleagues.

Meet the team, explore their research and join us beyond the lab, from picnics to mountain trails.

Meet the team Life at STAI

Take the next step

Start your next chapter at STAI.

Explore undergraduate internships, MSc and PhD opportunities.

Apply to STAI Explore openings Contact us
Affiliations
STAI STAI KAIST Tübingen AI Center University of Tübingen IMPRS-IS ELLIS