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

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Sennrich, Rico, Vamvas, Jannis, Mohammadshahi, Alireza
Format: Preprint
Published: 2023
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914655022088192
author Sennrich, Rico
Vamvas, Jannis
Mohammadshahi, Alireza
author_facet Sennrich, Rico
Vamvas, Jannis
Mohammadshahi, Alireza
contents 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.
format Preprint
id arxiv_https___arxiv_org_abs_2309_07098
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle Mitigating Hallucinations and Off-target Machine Translation with Source-Contrastive and Language-Contrastive Decoding
Sennrich, Rico
Vamvas, Jannis
Mohammadshahi, Alireza
Computation and Language
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.
title Mitigating Hallucinations and Off-target Machine Translation with Source-Contrastive and Language-Contrastive Decoding
topic Computation and Language
url https://arxiv.org/abs/2309.07098