Do World Models Learn Global Understanding?

Alexander Detkov, Matt Thomson

AI systems often feel brittle and fragmented. A large language model (LLM) may correctly explain a concept but fail to apply it, or follow safety instructions in one context but no…