Scientific Agents and Executable Evaluation
Series · 1 posts
-
ScienceIDE: Scientific Code as Verifiable Agent Environments
Advanced Agent runtime, safety, and evaluationA critical reading of ScienceIDE's 64 environments, 2,812 generated tasks, and separate 85-task hard benchmark: how reusable scientific checks support evaluation and training, and where private verifiers, partial release, task selection, and transfer claims remain bounded.
Understand it in 90 seconds
For speaking invitations, internal engineering sessions, or architecture exchange, see the topics and public work I can bring into the conversation.
Speaking & contact