ObjectConcepts

Study: Multimodal LLMs spontaneously develop human-like object concept representations, revealing 66 dimensions aligned with brain patterns for next-gen AI cognition.