Why Exposure Bias Matters: An Imitation Learning Perspective of Error Accumulation in Language Generation

By Kushal Arora, Layla El Asri, Hareesh Bahuleyan, Jackie Chi Kit Cheung. 2022

131 citations recorded by the source corpus.

Current language generation models suffer from issues such as repetition, incoherence, and hallucinations. An often-repeated hypothesis for this brittleness of generation models is that it 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 during generation, analyze why perplexity fails to capture this accumulation of errors, and empirically show that this accumulation results in poor generation quality.

Subfields: Artificial Intelligence, Natural Language Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Natural Language Processing

Citations in the corpus

Is Reinforcement Learning (Not) for Natural Language Processing?: Benchmarks, Baselines, and Building Blocks for Natural Language Policy Optimization (2022, background): “However, these algorithms are known to suffer from exposure bias in generation (Chiang & Chen, 2021; Arora et al., 2022) and the cliff MDP problem (Huszár, 2015; Agarwal et al.”
On-Policy Distillation of Language Models: Learning from Self-Generated Mistakes (2023, background): “While He et al. (2019) argue that this mismatch may not be critical, several papers demonstrate that exposure bias leads to poor text generation (Zhang et al., 2019; Chiang & Chen, 2021; Arora et al., 2022).”
How Language Model Hallucinations Can Snowball (2023, background): “Finally, while our experiments have focused on simple multi-step problems that are suitable for breaking down step-by-step, we hypothesize that hallucination snowballing appears in open-ended text generation more broadly, where one mistake in the generation triggers more (Arora et al., 2022).”