ARE LLMS CAPABLE OF NON-VERBAL REASONING?

Summary from the TLDR AI newsletter

Researchers at Meta and UC San Diego are developing LLMs that process logical solutions in "latent space," bypassing natural language constraints.