Analyzing and Mitigating Dataset Artifacts in Natural Language Inference Models Using ELECTRA
Abstract
This paper investigates the challenges posed by dataset artifacts in Natural Language Inference (NLI) models, focusing on ELECTRA, a state-of-the-art transformer model. Dataset artifacts such as hypothesis-only biases, lexical overlap issues, and frequent label imbalances significantly impact model generalization, leading to erroneous predictions. We propose and evaluate a range of strategies, including adversarial training, data augmentation, instance weighting, and artifact-aware regularization, to mitigate these issues. Extensive experimental results demonstrate up to a 6% improvement in robustness and generalization, providing valuable insights for creating artifact-resistant NLP models.
How to Cite This Article
Himanshu Joshi (2024). Analyzing and Mitigating Dataset Artifacts in Natural Language Inference Models Using ELECTRA . International Journal of Multidisciplinary Research and Growth Evaluation (IJMRGE), 5(6), 1279-1286. DOI: https://doi.org/10.54660/.IJMRGE.2024.5.6.1279-1286