Emergent Specialization: Rare Token Neurons in Language Models

Jing Liu, Haozheng Wang, Yueheng Li

Keywords: mechanistic interpretability, large language model, training dynamic, power law

TL;DR: Specialized "rare token neurons" in large language models develop a distinctive three-phase organizational structure during training

Abstract: Large language models struggle with representing and generating rare tokens despite their importance in specialized domains. In this study, we identify neuron structures with exceptionally strong influence on language model’s prediction of rare tokens, termed as rare token neurons, and investigate the mechanism for their emergence and behavior. These neurons exhibit a characteristic three-phase organization (plateau, power-law, and rapid decay) that emerges dynamically during training, evolving from a homogeneous initial state to a functionally differentiated architecture. In the activation space, rare token neurons form a coordinated subnetwork that selectively co-activates while avoiding co-activation with other neurons. This functional specialization potentially correlates with the development of heavy-tailed weight distributions, suggesting a statistical mechanical basis for emergent specialization.

Student Paper: Yes

Submission Number: 94