Recent advancements in large language models (LLMs) have enabled in-context learning (ICL)-based methods that significantly outperform fine-tuning approaches for text-to-SQL tasks. However, their performance is still considerably lower than that of human experts on benchmarks that include complex schemas and queries, such as BIRD. This study considers the sensitivity of LLMs to the prompts and introduces a novel approach that leverages multiple prompts to explore a broader search space for possible answers and effectively aggregate them. Specifically, we robustly refine the database schema through schema linking using multiple prompts. Thereafter, we generate various candidate SQL queries based on the refined schema and diverse prompts. Finally, the candidate queries are filtered based on their confidence scores, and the optimal query is obtained through a multiple-choice selection that is presented to the LLM. When evaluated on the BIRD and Spider benchmarks, the proposed method achieved execution accuracies of 65.5% and 89.6%, respectively, significantly outperforming previous ICL-based methods.

In this paper, we describe Dunamu ML’s submission to the Financial Misinformation Detection (FMD) 2025 shared task. To address the low-resource challenge in FMD, we augmented a general domain misinformation detection dataset for training. We first collected claims, contexts, and misinformation labels from a public dataset. Then, we generated evidence for each label based on a closed LLM with few-shot examples extracted from the FMD training dataset. Finally, we oversampled the training data specific to the financial domain and augmented it with the generated data to perform supervised fine-tuning (SFT) on the LLM. When evaluated on the blind test dataset, our model achieved an F1 score of 84.67 in misinformation classification and a ROUGE-1 score of 81.21 in evidence generation, ranking first on the leaderboard in both aspects.

This paper presents the Dunamu-ml’s submission to the AVERITEC shared task of the 7th the Fact Extraction and VERification (FEVER) workshop. The task focused on discriminating whether each claim is a fact or not. Our method is powered by the combination of an LLM and a non-parametric lexicon-based method (i.e. BM25). Essentially, we augmented the list of evidences containing the query and the corresponding answers using an powerful LLM, then, retrieved the relative documents using the generated evidences. As such, our method made a great improvement over the baseline results, achieving 0.33 performance gain over the baseline in AveriTec score.

We present IntelliCAT, an interactive translation interface with neural models that streamline the post-editing process on machine translation output. We leverage two quality estimation (QE) models at different granularities: sentence-level QE, to predict the quality of each machine-translated sentence, and word-level QE, to locate the parts of the machine-translated sentence that need correction. Additionally, we introduce a novel translation suggestion model conditioned on both the left and right contexts, providing alternatives for specific words or phrases for correction. Finally, with word alignments, IntelliCAT automatically preserves the original document’s styles in the translated document. The experimental results show that post-editing based on the proposed QE and translation suggestions can significantly improve translation quality. Furthermore, a user study reveals that three features provided in IntelliCAT significantly accelerate the post-editing task, achieving a 52.9% speedup in translation time compared to translating from scratch. The interface is publicly available at https://intellicat.beringlab.com/.

Automated metaphor detection is a challenging task to identify the metaphorical expression of words in a sentence. To tackle this problem, we adopt pre-trained contextualized models, e.g., BERT and RoBERTa. To this end, we propose a novel metaphor detection model, namely metaphor-aware late interaction over BERT (MelBERT). Our model not only leverages contextualized word representation but also benefits from linguistic metaphor identification theories to detect whether the target word is metaphorical. Our empirical results demonstrate that MelBERT outperforms several strong baselines on four benchmark datasets, i.e., VUA-18, VUA-20, MOH-X, and TroFi.

This paper presents the Bering Lab’s submission to the shared tasks of the 8th Workshop on Asian Translation (WAT 2021) on JPC2 and NICT-SAP. We participated in all tasks on JPC2 and IT domain tasks on NICT-SAP. Our approach for all tasks mainly focused on building NMT systems in domain-specific corpora. We crawled patent document pairs for English-Japanese, Chinese-Japanese, and Korean-Japanese. After cleaning noisy data, we built parallel corpus by aligning those sentences with the sentence-level similarity scores. Also, for SAP test data, we collected the OPUS dataset including three IT domain corpora. We then trained transformer on the collected dataset. Our submission ranked 1st in eight out of fourteen tasks, achieving up to an improvement of 2.87 for JPC2 and 8.79 for NICT-SAP in BLEU score .