🤖 AI Summary
This study investigates whether large language models spontaneously develop brain-like functional specialization to support high-level cognition. By applying integrated information decomposition, the authors systematically quantify information synergy and redundancy across layers in diverse architectures. They report the first evidence that intermediate layers undergo a phase-transition-like emergence of highly synergistic cores during complex tasks, while early and late layers predominantly rely on redundant processing, collectively forming a dynamic, brain-inspired organization. Ablation experiments demonstrate that this synergistic structure is critical for abstract reasoning, as its removal leads to catastrophic performance degradation, thereby establishing it as a core physical substrate of model intelligence. These findings offer novel empirical support for bridging artificial and biological intelligence.
📝 Abstract
The evolution of intelligence in artificial systems provides a unique opportunity to identify universal computational principles. Here we show that large language models spontaneously develop synergistic cores where information integration exceeds individual parts remarkably similar to the human brain. Using Integrated Information Decomposition across multiple architectures we find that middle layers exhibit synergistic processing while early and late layers rely on redundancy. This organization is dynamic and emerges as a physical phase transition as task difficulty increases. Crucially ablating synergistic components causes catastrophic performance loss confirming their role as the physical entity of abstract reasoning and bridging artificial and biological intelligence.