A tutorial series exploring how large language models can generate scientific hypotheses, learn to search, and turn experimental evidence into new knowledge.
Explore the tutorialsScaling Training, Hypothesis Coverage,
and Closed-Loop Experimentation
Beyond prompting: train discovery models, explore the hypothesis space, and let experiments guide what comes next.
Explore AAAI 2027From plausible ideas
to testable discoveries.
AAAI · Singapore
Benchmarks, Frameworks, and Applications
January 21, 2026 · Original tutorial page, slides, and reading list.
About the series
Scientific discovery asks more of AI than a plausible answer. It calls for hypotheses that can be tested, search that stays diverse, and systems that learn from evidence.
Our tutorials bring together methods, benchmarks, and demonstrations for AI researchers, graduate students, and domain scientists. Each edition offers a new perspective on the field while keeping earlier materials available to learn from.
Explore the latest tutorial