arXiv:2611.26228v1 Announce Type: new
Abstract: Attention heads specialize into distinct functions during training, often emerging in stages together with sharp changes in model behavior. While this phenomenon has mostly been studied as a function of training time, we ask how head specialization depends on the amount of available training data. Empirically, we find that heads specialized to syntactic relations in BERT emerge as the dataset size increases. We then study this phenomenon in a solvable high-dimensional model of multi-head attention and establish that, in this model, head specialization is a genuine phase transition in sample complexity, corresponding to a non-analyticity in the test error and other observables. Depending on data structure and regularization, specialization can follow uninformative head differentiation or occur through a cascade of phase transitions that progressively recover weaker latent features, providing a statistical form of simplicity bias. Finally, we relate these statistical transitions to the stagewise emergence of specialized heads during online learning, disentangling statistical from dynamical origins of head specialization.
Trending
- Scalable fault vulnerability analysis for ReRAM-based process-in-memory CNN accelerators
- This Is The Best Type Of Wireless Earbuds For Audio Output
- Attention Head Specialization: Phase Transitions in Sample Complexity
- [2611.26264] A Statistical Framework for Training Data Auditing of Large Language Models
- Building Your Own Mini PC Is Harder Than You Would Think
- [2611.26201] Learn, Freeze, and Transfer an Activation Function Across Architectures
- [2611.26235] Physics-Informed Feature Representations and Pretrained Tabular Foundation Models for Generalizable Melt Pool Geometry Prediction
- Invertible Transport Between Sample-Only Distributions via Projected Characteristic Functions
