

























































































Collaborators
ProjectsFrom benchmarks to
From benchmarks to
scientific superintelligence
AI agents run closed-loop experimental campaigns from real laboratories in biology, chemistry, and materials, designing each round of experiments and learning from the results.
Explore the bench
Cost vs. resolution rateAutonomous Science Bench 0.1
Fields
The science we cover






















Drag to explore
What we build
Built for scientific superintelligence
Benchmarks
Measure agents on real lab campaigns.
ASB 0.1 · 9 agent setups · 6 tasks
Environments
Train in the loop, replay to physical lab.
- v0Replay oracle
- v1Digital twin
- v2Physical lab
Same tasks · same agent API
Trajectories
Campaigns condensed into training tokens.
Models · simulations · experts · labs
Join us
Build scientific superintelligence with us
- For scientistsContribute a task that mattersPropose a task
- For researchers & engineersTurn scientific campaigns into scientific superintelligenceRun the bench
Join the teamyibo@u.northwestern.edu

