|国家预印本平台
首页|Parameter-Efficient Transformer Embeddings

Parameter-Efficient Transformer Embeddings

Parameter-Efficient Transformer Embeddings

来源:Arxiv_logoArxiv
英文摘要

Embedding layers in transformer-based NLP models typically account for the largest share of model parameters, scaling with vocabulary size but not yielding performance gains proportional to scale. We propose an alternative approach in which token embedding vectors are first generated deterministically, directly from the token IDs using a Fourier expansion of their normalized values, followed by a lightweight multilayer perceptron (MLP) that captures higher-order interactions. We train standard transformers and our architecture on natural language inference tasks (SNLI and MNLI), and evaluate zero-shot performance on sentence textual similarity (STS-B). Our results demonstrate that the proposed method achieves competitive performance using significantly fewer parameters, trains faster, and operates effectively without the need for dropout. This proof-of-concept study highlights the potential for scalable, memory-efficient language models and motivates further large-scale experimentation based on our findings.

Henry Ndubuaku、Mouad Talhi

计算技术、计算机技术

Henry Ndubuaku,Mouad Talhi.Parameter-Efficient Transformer Embeddings[EB/OL].(2025-05-04)[2025-07-23].https://arxiv.org/abs/2505.02266.点此复制

评论