Current language generation models suffer from issues such as repetition, incoherence, and hallucinations. An often-repeated hypothesis is that this brittleness of generation models is caused by the training and the generation procedure mismatch, also referred to as exposure bias. In this paper, we verify this hypothesis by analyzing exposure bias from an imitation learning perspective. We show that exposure bias leads to an accumulation of errors, analyze why perplexity fails to capture this accumulation, and empirically show that this accumulation results in poor generation quality. Source code to reproduce these experiments is available at https://github.com/kushalarora/quantifying_exposure_bias
@article{arxiv.2204.01171,
title = {Why Exposure Bias Matters: An Imitation Learning Perspective of Error Accumulation in Language Generation},
author = {Kushal Arora and Layla El Asri and Hareesh Bahuleyan and Jackie Chi Kit Cheung},
journal= {arXiv preprint arXiv:2204.01171},
year = {2023}
}
Comments
Accepted in Findings of ACL 2022. v2: Equation 7 updated, typo fixes