SplaXBERT: Leveraging Mixed Precision Training and Context Splitting for Question Answering

TL;DR

SplaXBERT enhances question-answering on lengthy texts by combining context splitting and mixed precision training, achieving superior accuracy and efficiency on SQuAD v1.1 compared to traditional BERT models.

Contribution

This paper introduces SplaXBERT, a novel model that integrates context splitting with mixed precision training to improve performance and resource efficiency in question-answering tasks.

Findings

Achieves 85.95% Exact Match on SQuAD v1.1

Attains 92.97% F1 Score, outperforming traditional models

Demonstrates improved efficiency in processing lengthy texts

Abstract

SplaXBERT, built on ALBERT-xlarge with context-splitting and mixed precision training, achieves high efficiency in question-answering tasks on lengthy texts. Tested on SQuAD v1.1, it attains an Exact Match of 85.95% and an F1 Score of 92.97%, outperforming traditional BERT-based models in both accuracy and resource efficiency.