3 projects
vlabs-calibrate
Conformal coverage guarantees for any reward function — wrap a Python callable with provable (1-α) coverage in 5 lines.
verifiable-labs
Python SDK + CLI for Verifiable Labs — evaluate frontier LLMs on conformal-calibrated scientific RL environments.
verifiable-labs-envs
Reinforcement-learning environments for scientific reasoning: physics-grounded inverse problems with uncertainty-calibrated rewards.
