Apr 21
Defragmenting Language Models: An Interpretability-based Approach for Vocabulary Expansion
★★★★★
significance 3/5
The paper investigates 'token over-fragmentation' in large language models, where non-Latin scripts require more tokens than English. It proposes a new interpretability-based approach for vocabulary expansion and embedding initialization to improve efficiency and performance.
Why it matters
Moving beyond frequency-based expansion addresses the fundamental efficiency bottlenecks and linguistic biases inherent in current LLM tokenization architectures.
Tags
#tokenization #interpretability #llm #vocabulary expansion #nlpRelated coverage
- Global South OpportunitiesPivotal Research Fellowship 2026 (Q3): AI Safety Research Opportunity - Global South Opportunities
- arXiv cs.AIAn Intelligent Fault Diagnosis Method for General Aviation Aircraft Based on Multi-Fidelity Digital Twin and FMEA Knowledge Enhancement
- arXiv cs.AIPExA: Parallel Exploration Agent for Complex Text-to-SQL
- arXiv cs.AIThe Power of Power Law: Asymmetry Enables Compositional Reasoning
- arXiv cs.AIOn the Existence of an Inverse Solution for Preference-Based Reductions in Argumentation