CMU's Sim2Reason trains language models inside physics simulations rather than on static text, which improved their performance on real-world physics reasoning tasks.

• articlePopular AI Models Aren’t Ready to Safely Power Robots
• articleHealthcare Blind Spots: AI Models Prone To Fabricating Diagnoses

• articleSystem Modeling: The Substrate for Neural Co-EvolutionCarlos H. Morales
• articleSimulation: the new Scaling Law — Joon Sung Park, Simile AIwww.latent.space
• article🔬“We have foundation models for language, not for physics” — Anima Anandkumar, Bren Professor of ComputingBrandon Anderson

Checking sign-in…

Loading comments…