Unraveling the cognitive patterns of Large Language Models through module communities
Abstract
A network-based framework links cognitive skills, LLM architectures, and datasets, revealing unique emergent skill patterns in LLMs that benefit from dynamic, cross-regional interactions.
Large Language Models (LLMs) have reshaped our world with significant advancements in science, engineering, and society through applications ranging from scientific discoveries and medical diagnostics to Chatbots. Despite their ubiquity and utility, the underlying mechanisms of LLM remain concealed within billions of parameters and complex structures, making their inner architecture and cognitive processes challenging to comprehend. We address this gap by adopting approaches to understanding emerging cognition in biology and developing a network-based framework that links cognitive skills, LLM architectures, and datasets, ushering in a paradigm shift in foundation model analysis. The skill distribution in the module communities demonstrates that while LLMs do not strictly parallel the focalized specialization observed in specific biological systems, they exhibit unique communities of modules whose emergent skill patterns partially mirror the distributed yet interconnected cognitive organization seen in avian and small mammalian brains. Our numerical results highlight a key divergence from biological systems to LLMs, where skill acquisition benefits substantially from dynamic, cross-regional interactions and neural plasticity. By integrating cognitive science principles with machine learning, our framework provides new insights into LLM interpretability and suggests that effective fine-tuning strategies should leverage distributed learning dynamics rather than rigid modular interventions.
Community
The paper explores how large language models (LLMs) organize and express cognitive skills by mapping datasets, skills, and the model's weight modules into a multipartite network. Using network analysis and pruning techniques, the authors demonstrate that LLMs comprise communities of modules that partially mirror the distributed yet interconnected organization of animal brains, albeit without the same strict specialization. Fine-tuning skill-specific modules changes weights significantly but does not improve accuracy compared to random modules, suggesting that LLM capabilities arise from broad, redundant, and interdependent structures rather than isolated components. This framework blends insights from network science and cognitive science to improve interpretability and points toward future strategies that leverage network-wide dynamics for fine-tuning and optimization
This is an automated message from the Librarian Bot. I found the following papers similar to this paper.
The following papers were recommended by the Semantic Scholar API
- Pruning Large Language Models by Identifying and Preserving Functional Networks (2025)
- The Generalization Ridge: Information Flow in Natural Language Generation (2025)
- NDAI-NeuroMAP: A Neuroscience-Specific Embedding Model for Domain-Specific Retrieval (2025)
- From Scores to Skills: A Cognitive Diagnosis Framework for Evaluating Financial Large Language Models (2025)
- The Prompting Brain: Neurocognitive Markers of Expertise in Guiding Large Language Models (2025)
- An Empirical Study of Knowledge Distillation for Code Understanding Tasks (2025)
- The Other Mind: How Language Models Exhibit Human Temporal Cognition (2025)
Please give a thumbs up to this comment if you found it helpful!
If you want recommendations for any Paper on Hugging Face checkout this Space
You can directly ask Librarian Bot for paper recommendations by tagging it in a comment:
@librarian-bot
recommend
Models citing this paper 0
No model linking this paper
Datasets citing this paper 0
No dataset linking this paper
Spaces citing this paper 0
No Space linking this paper