Summary Background A new autoinflammatory syndrome related to somatic mutations of UBA1 was recently described and called VEXAS syndrome (‘Vacuoles, E1 Enzyme, X‐linked, Autoinflammatory, Somatic syndrome’). Objectives To describe clinical characteristics, laboratory findings and outcomes of VEXAS syndrome. Methods One hundred and sixteen patients with VEXAS syndrome were referred to a French multicentre registry between November 2020 and May 2021. The frequency and median of parameters and vital status, from diagnosis to the end of the follow‐up, were recorded. Results The main clinical features of VEXAS syndrome were found to be skin lesions (83%), noninfectious fever (64%), weight loss (62%), lung involvement (50%), ocular symptoms (39%), relapsing chondritis (36%), venous thrombosis (35%), lymph nodes (34%) and arthralgia (27%). Haematological disease was present in 58 cases (50%): myelodysplastic syndrome (MDS; n = 58) and monoclonal gammopathy of unknown significance (n = 12; all patients with MGUS also have a MDS). UBA1 mutations included p.M41T (45%), p.M41V (30%), p.M41L (18%) and splice mutations (7%). After a median follow‐up of 3 years, 18 patients died (15·5%; nine of infection and three due to MDS progression). Unsupervised analysis identified three clusters: cluster 1 (47%; mild‐to‐moderate disease); cluster 2 (16%; underlying MDS and higher mortality rates); and cluster 3 (37%; constitutional manifestations, higher C‐reactive protein levels and less frequent chondritis). The 5‐year probability of survival was 84·2% in cluster 1, 50·5% in cluster 2 and 89·6% in cluster 3. The UBA1 p.Met41Leu mutation was associated with a better prognosis. Conclusions VEXAS syndrome has a large spectrum of organ manifestations and shows different clinical and prognostic profiles. It also raises a potential impact of the identified UBA1 mutation.

Background Fat infiltration in individual muscles of sporadic inclusion body myositis (sIBM) patients has rarely been assessed. Methods Sixteen sIBM patients were assessed using MRI of the thighs and lower legs (LL). The severity of fat infiltration, proximal‐to‐distal and side asymmetries, and the correlations with clinical and functional parameters were investigated. Results All the patients had fat‐infiltrated muscles, and thighs were more severely affected than LL. A proximal‐to‐distal gradient of fat infiltration was mainly observed for adductors, quadriceps, sartorius, and medial gastrocnemius muscles. A strong negative correlation was observed between the whole muscle fat fraction in the thighs and LL and the Inclusion Body Myositis Functional Rating Scale and Medical Research Council scores for the lower limbs. Conclusions Fat infiltration in individual muscles of sIBM patients is heterogeneous in terms of proximal‐to‐distal gradient and severity was correlated with clinical scores. These results should be considered for both natural history investigation and clinical trials.

In this paper, we show that neural machine translation (NMT) systems trained on large back-translated data overfit some of the characteristics of machine-translated texts. Such NMT systems better translate humanproduced translations, i.e., translationese, but may largely worsen the translation quality of original texts. Our analysis reveals that adding a simple tag to back-translations prevents this quality degradation and improves on average the overall translation quality by helping the NMT system to distinguish back-translated data from original parallel data during training. We also show that, in contrast to high-resource configurations, NMT systems trained in lowresource settings are much less vulnerable to overfit back-translations. We conclude that the back-translations in the training data should always be tagged especially when the origin of the text to be translated is unknown.

This paper presents the first large-scale metaevaluation of machine translation (MT). We annotated MT evaluations conducted in 769 research papers published from 2010 to 2020. Our study shows that practices for automatic MT evaluation have dramatically changed during the past decade and follow concerning trends. An increasing number of MT evaluations exclusively rely on differences between BLEU scores to draw conclusions, without performing any kind of statistical significance testing nor human evaluation, while at least 108 metrics claiming to be better than BLEU have been proposed. MT evaluations in recent papers tend to copy and compare automatic metric scores from previous work to claim the superiority of a method or an algorithm without confirming neither exactly the same training, validating, and testing data have been used nor the metric scores are comparable. Furthermore, tools for reporting standardized metric scores are still far from being widely adopted by the MT community. After showing how the accumulation of these pitfalls leads to dubious evaluation, we propose a guideline to encourage better automatic MT evaluation along with a simple meta-evaluation scoring method to assess its credibility.