Mitigating Hallucinations and Off-target Machine Translation with Source-Contrastive and Language-Contrastive Decoding

TL;DR

This paper presents source-contrastive and language-contrastive decoding methods that significantly reduce hallucinations and off-target translations in multilingual machine translation models without retraining.

Contribution

It introduces two decoding strategies that mitigate hallucinations and off-target issues in multilingual MT, applicable without retraining or external models.

Findings

Reduces segment-level chrF2 below 10 by 67-83%

Decreases oscillatory hallucinations by 75-92%

Effective in multilingual and out-of-English translation scenarios

Abstract

Hallucinations and off-target translation remain unsolved problems in MT, especially for low-resource languages and massively multilingual models. In this paper, we introduce two related methods to mitigate these failure cases with a modified decoding objective, without either requiring retraining or external models. In source-contrastive decoding, we search for a translation that is probable given the correct input, but improbable given a random input segment. In language-contrastive decoding, we search for a translation that is probable, but improbable given the wrong language indicator token. Experiments on the massively multilingual models M2M-100 (418M) and SMaLL-100 show that these methods suppress hallucinations and off-target translations, reducing the number of translations with segment-level chrF2 below 10 by 67-83% on average, and the number of translations with oscillatory hallucinations by 75-92% on average, across 57 tested translation directions. In a proof of concept on out-of-English translation, we also show that we can suppress off-target translations with large language models. We release our source code at https://github.com/ZurichNLP/ContraDecode.