Verification Infrastructure for Autonomous Science

Endless Worlds

We generate training and evaluation environments where scientific agents learn to reason, verify, and discover.

Research

Scaling Test-Time Verification for Novel Materials

Crystal diffusion models encode property signals in their hidden states that they never use during sampling. Probe-gradient guidance steers unconditional generation toward target properties at test time, comparable to conditional models at over 50x the speed.

BlogApril 2026

The Missing Layer in Autonomous Science

Verified campaign environments with physics-grounded oracles provide the missing training signal for sequential decision-making in autonomous scientific workflows.

BlogApril 2026