Human-like object concept representations emerge naturally in multimodal large language models

2025/06/09 by Changde Du, Kaicheng Fu, Bin Wen +11 · 1 voice · 3 citations
Computer Science · #Natural Language Processing Techniques #Speech and dialogue systems #Topic Modeling

paper · doi:10.1038/s42256-025-01049-z

openalex created_date 2024/07/06 · crossref issued 2025/06/09 · crossref published 2025/06/09 · crossref published-online 2025/06/09 · openalex publication_date 2025/06/09 · crossref created 2025/06/09 · crossref deposited 2025/06/23 · crossref indexed 2026/07/16 · openalex updated_date 2026/07/28

Citations

• Representation is representation of similarities
• Attention, similarity, and the identification–categorization relationship.
• Basic objects in natural categories
• Deep problems with neural network models of human vision
• Testing theory of mind in large language models and humans
• A massive 7T fMRI dataset to bridge cognitive neuroscience and artificial intelligence
• Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolution
• Learning Transferable Visual Models From Natural Language Supervision
• From attribution maps to human-understandable explanations through Concept Relevance Propagation

Cited by

• iMIND: Insightful Multi-subject Invariant Neural Decoding
• VeriOS: Query-Driven Proactive Human-Agent-GUI Interaction for Trustworthy OS Agents
• Language models align with brain regions that represent concepts across modalities

Discussions

• doi.org/10.1038/s422... [bsky, 1 points, 0 comments]