Les systèmes de questions-réponses (QA) actuels ont du mal à synthétiser les preuves dispersées dans les documents. Alors que les jeux de données QA scientifiques existants se concentrent sur le raisonnement portant sur un document seul, la tâche de recherche peut exiger l’intégration de contenus provenant de plusieurs articles. Pour répondre à cette limitation, nous proposons un cadre pour créer un jeu de données QA multi-documents qui s’appuie sur l’analyse de graphes de citations afin de regrouper des articles connexes et utilise un grand modèle de langage (LLM) pour générer des questions complexes. Des expériences préliminaires réalisées sur 23 882 articles démontrent la faisabilité de ce cadre, produisant 238 paires QA qui nécessitent une synthèse sur plusieurs articles. D’autres expériences indiquent que la recherche d’information dense actuelle obtient un rappel limité pour ces questions multi-documents, soulignant le besoin de mécanismes de recherche d’information et de raisonnement plus avancés. Il s’agit d’un projet en cours d’élaboration. Nous visons à terme à fournir un jeu de données QA robuste qui capture la complexité et la nature interconnectée des publications scientifiques, ouvrant la voie à des évaluations plus réalistes des systèmes de QA.

This paper describes our submission for SemEval-2024 Task 6: SHROOM, a Shared-task on Hallucinations and Related Observable Overgeneration Mistakes. We propose four groups of methods for hallucination detection: 1) Entailment Recognition; 2) Similarity Search; 3) Factuality Verification; 4) Confidence Estimation. The four methods rely on either the semantic relationship between the hypothesis and its source (target) or on the model-aware features during decoding. We participated in both the model-agnostic and model-aware tracks. Our method’s effectiveness is validated by our high rankings 3rd in the model-agnostic track and 5th in the model-aware track. We have released our code on GitHub.

The proliferation of open-source Large Language Models (LLMs) underscores the pressing need for evaluation methods. Existing works primarily rely on external evaluators, focusing on training and prompting strategies. However, a crucial aspect – model-aware glass-box features – is overlooked. In this study, we explore the utility of glass-box features under the scenario of self-evaluation, namely applying an LLM to evaluate its own output. We investigate various glass-box feature groups and discovered that the softmax distribution serves as a reliable quality indicator for self-evaluation. Experimental results on public benchmarks validate the feasibility of self-evaluation of LLMs using glass-box features.

Recently, there has been a trend of evaluating the Large Language Model (LLM) quality in the flavor of LLM-as-a-Judge, namely leveraging another LLM to evaluate the current output quality. However, existing judges are proven to be biased, namely they would favor answers which present better superficial quality (such as verbosity, fluency) while ignoring the instruction following ability. In this work, we propose systematic research about the bias of LLM-as-a-Judge. Specifically, for closed-source judge models, we apply calibration to mitigate the significance of superficial quality, both on probability level and prompt level. For open-source judge models, we propose to mitigate the bias by contrastive training, with curated negative samples that deviate from instruction but present better superficial quality. We apply our methods on the bias evaluation benchmark, and experiment results show our methods mitigate the bias by a large margin while maintaining a satisfactory evaluation accuracy.

Recently, Large Language Models (LLMs) have boosted the research in natural language processing and shown impressive capabilities across numerous domains, including machine translation evaluation. This paper presents our methods developed for the machine translation evaluation sub-task of the Eval4NLP 2023 Shared Task. Based on the provided LLMs, we propose a generation-based method as well as a probability-based method to perform evaluation, explore different strategies when selecting the demonstrations for in-context learning, and try different ensemble methods to further improve the evaluation accuracy. The experiment results on the development set and test set demonstrate the effectiveness of our proposed method.

“Recent works in dialogue state tracking (DST) focus on a handful of languages, as collectinglarge-scale manually annotated data in different languages is expensive. Existing models addressthis issue by code-switched data augmentation or intermediate fine-tuning of multilingual pre-trained models. However, these models can only perform implicit alignment across languages. In this paper, we propose a novel model named Contrastive Learning for Cross-Lingual DST(CLCL-DST) to enhance zero-shot cross-lingual adaptation. Specifically, we use a self-builtbilingual dictionary for lexical substitution to construct multilingual views of the same utterance. Then our approach leverages fine-grained contrastive learning to encourage representations ofspecific slot tokens in different views to be more similar than negative example pairs. By thismeans, CLCL-DST aligns similar words across languages into a more refined language-invariantspace. In addition, CLCL-DST uses a significance-based keyword extraction approach to selecttask-related words to build the bilingual dictionary for better cross-lingual positive examples. Experiment results on Multilingual WoZ 2.0 and parallel MultiWoZ 2.1 datasets show that ourproposed CLCL-DST outperforms existing state-of-the-art methods by a large margin, demon-strating the effectiveness of CLCL-DST.”

This paper presents the BJTU-Toshiba joint submission for WMT 2022 quality estimation shared task. We only participate in Task 1 (quality prediction) of the shared task, focusing on the sentence-level MQM prediction. The techniques we experimented with include the integration of monolingual language models and the pre-finetuning of pre-trained representations. We tried two styles of pre-finetuning, namely Translation Language Modeling and Replaced Token Detection. We demonstrate the competitiveness of our system compared to the widely adopted XLM-RoBERTa baseline. Our system is also the top-ranking system on the Sentence-level MQM Prediction for the English-German language pairs.

Translation suggestion (TS) models are used to automatically provide alternative suggestions for incorrect spans in sentences generated by machine translation. This paper introduces the system used in our submission to the WMT’22 Translation Suggestion shared task. Our system is based on the ensemble of different translation architectures, including Transformer, SA-Transformer, and DynamicConv. We use three strategies to construct synthetic data from parallel corpora to compensate for the lack of supervised data. In addition, we introduce a multi-phase pre-training strategy, adding an additional pre-training phase with in-domain data. We rank second and third on the English-German and English-Chinese bidirectional tasks, respectively.

Back-translation has been proven to be effective in unsupervised domain adaptation of neural machine translation (NMT). However, the existing back-translation methods mainly improve domain adaptability by generating in-domain pseudo-parallel data that contains sentence-structural knowledge, paying less attention to the in-domain lexical knowledge, which may lead to poor translation of unseen in-domain words. In this paper, we propose an Iterative Constrained Back-Translation (ICBT) method to incorporate in-domain lexical knowledge on the basis of BT for unsupervised domain adaptation of NMT. Specifically, we apply lexical constraints into back-translation to generate pseudo-parallel data with in-domain lexical knowledge, and then perform round-trip iterations to incorporate more lexical knowledge. Based on this, we further explore sampling strategies of constrained words in ICBT to introduce more targeted lexical knowledge, via domain specificity and confidence estimation. Experimental results on four domains show that our approach achieves state-of-the-art results, improving the BLEU score by up to 3.08 compared to the strongest baseline, which demonstrates the effectiveness of our approach.

Recent multilingual pre-trained models, like XLM-RoBERTa (XLM-R), have been demonstrated effective in many cross-lingual tasks. However, there are still gaps between the contextualized representations of similar words in different languages. To solve this problem, we propose a novel framework named Multi-View Mixed Language Training (MVMLT), which leverages code-switched data with multi-view learning to fine-tune XLM-R. MVMLT uses gradient-based saliency to extract keywords which are the most relevant to downstream tasks and replaces them with the corresponding words in the target language dynamically. Furthermore, MVMLT utilizes multi-view learning to encourage contextualized embeddings to align into a more refined language-invariant space. Extensive experiments with four languages show that our model achieves state-of-the-art results on zero-shot cross-lingual sentiment classification and dialogue state tracking tasks, demonstrating the effectiveness of our proposed model.