Understanding and Detecting Hallucinations in Neural Machine Translation via Model Introspection
This addresses the issue of harmful hallucinations in translation for users, but is incremental as it builds on existing detection methods.
The paper tackled the problem of hallucinations in neural machine translation by identifying internal model symptoms and using them to design a detector, which outperformed baselines on English-Chinese and German-English test beds.
Neural sequence generation models are known to "hallucinate", by producing outputs that are unrelated to the source text. These hallucinations are potentially harmful, yet it remains unclear in what conditions they arise and how to mitigate their impact. In this work, we first identify internal model symptoms of hallucinations by analyzing the relative token contributions to the generation in contrastive hallucinated vs. non-hallucinated outputs generated via source perturbations. We then show that these symptoms are reliable indicators of natural hallucinations, by using them to design a lightweight hallucination detector which outperforms both model-free baselines and strong classifiers based on quality estimation or large pre-trained models on manually annotated English-Chinese and German-English translation test beds.