Universal Neurons in GPT-2: Emergence, Persistence, and Functional Impact
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866911255177986048 |
|---|---|
| author | Nandan, Advey Chou, Cheng-Ting Kurakula, Amrit Blondin, Cole Zhu, Kevin Sharma, Vasu O'Brien, Sean |
| author_facet | Nandan, Advey Chou, Cheng-Ting Kurakula, Amrit Blondin, Cole Zhu, Kevin Sharma, Vasu O'Brien, Sean |
| contents | We investigate the phenomenon of neuron universality in independently trained GPT-2 Small models, examining these universal neurons-neurons with consistently correlated activations across models-emerge and evolve throughout training. By analyzing five GPT-2 models at five checkpoints, we identify universal neurons through pairwise correlation analysis of activations over a dataset of 5 million tokens. Ablation experiments reveal significant functional impacts of universal neurons on model predictions, measured via cross entropy loss. Additionally, we quantify neuron persistence, demonstrating high stability of universal neurons across training checkpoints, particularly in early and deeper layers. These findings suggest stable and universal representational structures emerge during language model training. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2508_00903 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Universal Neurons in GPT-2: Emergence, Persistence, and Functional Impact Nandan, Advey Chou, Cheng-Ting Kurakula, Amrit Blondin, Cole Zhu, Kevin Sharma, Vasu O'Brien, Sean Machine Learning Artificial Intelligence Neural and Evolutionary Computing We investigate the phenomenon of neuron universality in independently trained GPT-2 Small models, examining these universal neurons-neurons with consistently correlated activations across models-emerge and evolve throughout training. By analyzing five GPT-2 models at five checkpoints, we identify universal neurons through pairwise correlation analysis of activations over a dataset of 5 million tokens. Ablation experiments reveal significant functional impacts of universal neurons on model predictions, measured via cross entropy loss. Additionally, we quantify neuron persistence, demonstrating high stability of universal neurons across training checkpoints, particularly in early and deeper layers. These findings suggest stable and universal representational structures emerge during language model training. |
| title | Universal Neurons in GPT-2: Emergence, Persistence, and Functional Impact |
| topic | Machine Learning Artificial Intelligence Neural and Evolutionary Computing |
| url | https://arxiv.org/abs/2508.00903 |