We teach machines to read & write
— with people in mind.
We’re an interdisciplinary research group drawing from linguistics, the social sciences, and cognitive science to build state-of-the-art AI that supports expert work in law, science, education, and journalism.
Led by Dongyeop Kang & Alexander Spangher (joining Fall 2026) in Computer Science & Engineering at the University of Minnesota — Twin Cities.
Five questions we keep returning to.
Our research spans modeling and alignment through to cognition and the messy, expert workflows where language really matters.
Modeling
Controllable generation, structure-aware tuning, and efficient adaptation of large language models.
Alignment
Reward learning, preference modeling, and calibrated abstention so systems know what they don’t know.
Evaluation
LLM-as-judge bias, cognitive benchmarks, and principled assessment of model behavior.
Cognition
Reading processes, executive function, and eye-tracking signals that connect language and the mind.
Expert Workflows
Legal reasoning, scientific writing, and journalism support built alongside the people who do the work.
An interdisciplinary team.
Linguists, computer scientists, and cognitive scientists — advised by Dongyeop Kang, with Alexander Spangher joining in Fall 2026.
We take the work seriously — not ourselves.
Picnics, escape rooms, paper clinics, and the occasional tennis drill. A research group is a community first.
