Neural Activation Patterns Across Language Model Architectures: A Comprehensive Analysis of Cognitive Task Performance
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2026
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866911686318882816 |
|---|---|
| author | Naser-Moghadasi, Mahdi Ghaderi, Faezeh |
| author_facet | Naser-Moghadasi, Mahdi Ghaderi, Faezeh |
| contents | This paper presents a comprehensive analysis of neural activation patterns across six distinct large language model (LLM) architectures, examining their performance on twelve cognitive task categories. Through systematic measurement of final activation values, attention entropy, and sparsity patterns, we reveal fundamental differences in how encoder and decoder architectures process diverse cognitive tasks. Our analysis of 144 task-model combinations demonstrates that mathematical reasoning consistently produces the highest attention entropy across all architectures, while decoder models exhibit significantly higher sparsity patterns compared to encoder models. The findings provide critical insights into the computational characteristics of modern language models and their task-specific neural behaviors, with implications for model selection and optimization in big data applications. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_15436 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Neural Activation Patterns Across Language Model Architectures: A Comprehensive Analysis of Cognitive Task Performance Naser-Moghadasi, Mahdi Ghaderi, Faezeh Computation and Language Machine Learning I.2.7; I.2.6 This paper presents a comprehensive analysis of neural activation patterns across six distinct large language model (LLM) architectures, examining their performance on twelve cognitive task categories. Through systematic measurement of final activation values, attention entropy, and sparsity patterns, we reveal fundamental differences in how encoder and decoder architectures process diverse cognitive tasks. Our analysis of 144 task-model combinations demonstrates that mathematical reasoning consistently produces the highest attention entropy across all architectures, while decoder models exhibit significantly higher sparsity patterns compared to encoder models. The findings provide critical insights into the computational characteristics of modern language models and their task-specific neural behaviors, with implications for model selection and optimization in big data applications. |
| title | Neural Activation Patterns Across Language Model Architectures: A Comprehensive Analysis of Cognitive Task Performance |
| topic | Computation and Language Machine Learning I.2.7; I.2.6 |
| url | https://arxiv.org/abs/2605.15436 |