Single-Nodal Spontaneous Symmetry Breaking in NLP Models

TL;DR

This paper reveals that NLP models like BERT can exhibit spontaneous symmetry breaking at the single-node level during training, affecting their learning capacity and cooperation among nodes, with implications for understanding model behavior.

Contribution

It demonstrates the occurrence of spontaneous symmetry breaking in NLP models at the individual node level during training, a phenomenon previously known only in physical systems.

Findings

Symmetry breaking occurs at the level of individual attention heads.

Nodes acquire specific token or label learning capabilities post-training.

A crossover in learning ability emerges as the number of nodes increases.

Abstract

Spontaneous symmetry breaking in statistical mechanics primarily occurs during phase transitions at the thermodynamic limit where the Hamiltonian preserves inversion symmetry, yet the low-temperature free energy exhibits reduced symmetry. Herein, we demonstrate the emergence of spontaneous symmetry breaking in natural language processing (NLP) models during both pre-training and fine-tuning, even under deterministic dynamics and within a finite training architecture. This phenomenon occurs at the level of individual attention heads and is scaled-down to its small subset of nodes and also valid at a single-nodal level, where nodes acquire the capacity to learn a limited set of tokens after pre-training or labels after fine-tuning for a specific classification task. As the number of nodes increases, a crossover in learning ability occurs, governed by the tradeoff between a decrease following random-guess among increased possible outputs, and enhancement following nodal cooperation, which exceeds the sum of individual nodal capabilities. In contrast to spin-glass systems, where a microscopic state of frozen spins cannot be directly linked to the free-energy minimization goal, each nodal function in this framework contributes explicitly to the global network task and can be upper-bounded using convex hull analysis. Results are demonstrated using BERT-6 architecture pre-trained on Wikipedia dataset and fine-tuned on the FewRel classification task.