This paper presents our submission to Subtask 2 (multi-label classification of persuasion techniques) of the Shared Task on Detection and Classification of Persuasion Techniques in Slavic Languages at SlavNLP 2025. Our method leverages a teacher–student framework based on large language models (LLMs): a Qwen3 32B teacher model generates natural language explanations for annotated persuasion techniques, and a Qwen2.5 32B student model is fine-tuned to replicate both the teacher’s rationales and the final label predictions. We train our models on the official shared task dataset, supplemented by annotated resources from SemEval 2023 Task 3 and CLEF 2024 Task 3 covering English, Russian, and Polish to improve cross-lingual robustness. Our final system ranks 4th on BG, SI, and HR, and 5th on PL in terms of micro-F1 score among all participating teams.

Few-Shot Document-Level Relation Extraction (FSDLRE) aims to develop models capable of generalizing to new categories with minimal support examples. Although Large Language Models (LLMs) demonstrate exceptional In-Context Learning (ICL) capabilities on many few-shot tasks, their performance on FSDLRE tasks remains suboptimal due to the significant gap between the task format and the intrinsic capabilities of language models, coupled with the complexity of ICL prompts for document-level text. To address these challenges, we introduce a novel meta-training approach for LLMs termed Prototype Tuning. We construct simulated episodes using data with relation types that do not overlap with the test corpus, fundamentally enhancing the ICL capabilities of LLMs in FSDLRE through meta-learning. To further enhance the effects of meta-learning, we innovatively integrate the concept of prototype into the fine-tuning process of LLMs. This involves aggregating entity pairs from support documents into prototypes within the prompts and altering the way of determining relation categories to identifying the closest prototype. Experimental results demonstrate that our LLMs trained with this approach outperform all baselines. Our proposed approach markedly improves the ICL capabilities of LLMs in FSDLRE and mitigates the impact of relation semantic discrepancies between the training corpus and the test corpus on model performance.

Retrieval-augmented generation (RAG) has become a fundamental paradigm for addressing the challenges faced by large language models in handling real-time information and domain-specific problems. Traditional RAG systems primarily rely on the in-context learning (ICL) capabilities of the large language model itself. Still, in-depth research on the specific capabilities needed by the RAG generation model is lacking, leading to challenges with inconsistent document quality and retrieval system imperfections. Even the limited studies that fine-tune RAG generative models often lack a granular focus on RAG tasks or a deeper utilization of chain-of-thought processes. To address this, we propose that RAG models should possess three progressively hierarchical abilities (1) Filtering: the ability to select relevant information; (2) Combination: the ability to combine semantic information across paragraphs; and (3) RAG-specific reasoning: the ability to further process external knowledge using internal knowledge. Thus, we introduce our new RAG instruction fine-tuning method, Hierarchical-Thought Instruction-Tuning Retrieval-Augmented Generation (HIRAG) incorporates a “think before answering” strategy. This method enhances the model’s open-book examination capability by utilizing multi-level progressive chain-of-thought. Experiments show that the HIRAG training strategy significantly improves the model’s performance on datasets such as RGB, PopQA, MuSiQue, HotpotQA, and PubmedQA.

“从文本中自动识别人类表型概念对疾病分析具有重大意义。现存本体驱动的表型概念识别方法主要利用本体中概念名和同义词信息,并未充分考虑本体丰富信息。针对此问题,本文提出一种基于本体信息增强的人类表型概念识别方法,利用先进大语言模型进行数据增强,并设计本体向量增强的深度学习模型来提升概念识别性能。在GSC+和ID-68两个数据集上进行实验,结果表明本文提出方法能够利用本体丰富信息有效提升基线模型性能,取得了先进结果。”

Emotion Support Conversation (ESC) is a crucial application, which aims to reduce human stress, offer emotional guidance, and ultimately enhance human mental and physical well-being. With the advancement of Large Language Models (LLMs), many researchers have employed LLMs as the ESC models. However, the evaluation of these LLM-based ESCs remains uncertain. In detail, we first re-organize 2,801 role-playing cards from seven existing datasets to define the roles of the role-playing agent. Second, we train a specific role-playing model called ESC-Role which behaves more like a confused person than GPT-4. Third, through ESC-Role and organized role cards, we systematically conduct experiments using 14 LLMs as the ESC models, including general AI-assistant LLMs (e.g., ChatGPT) and ESC-oriented LLMs (e.g., ExTES-Llama). We conduct comprehensive human annotations on interactive multi-turn dialogues of different ESC models. The results show that ESC-oriented LLMs exhibit superior ESC abilities compared to general AI-assistant LLMs, but there is still a gap behind human performance. Moreover, to automate the scoring process for future ESC models, we developed ESC-RANK, which trained on the annotated data, achieving a scoring performance surpassing 35 points of GPT-4.

Detecting persuasive communication is an important topic in Natural Language Processing (NLP), as it can be useful in identifying fake information on social media. We have developed a system to identify applied persuasion techniques in text fragments across four languages: English, Bulgarian, North Macedonian, and Arabic. Our system uses data augmentation methods and employs an ensemble strategy that combines the strengths of both RoBERTa and DeBERTa models. Due to limited resources, we concentrated solely on task 1, and our solution achieved the top ranking in the English track during the official assessments. We also analyse the impact of architectural decisions, data constructionand training strategies.

“语音翻译的编码器需要同时编码语音中的声学和语义信息,单一的Fbank或Wav2vec2语音特征表征能力存在不足。本文通过分析人工的Fbank特征与自监督的Wav2vec2特征间的差异性,提出基于交叉注意力机制的声学特征融合方法,并探究了不同的自监督特征和融合方式,加强模型对语音中声学和语义信息的学习。结合越南语语音特点,以Fbank特征为主、Pitch特征为辅混合编码Fbank表征,构建多特征融合的越-英语音翻译模型。实验表明,使用多特征的语音翻译模型相比单特征翻译效果更优,与简单的特征拼接方法相比更有效,所提的多特征融合方法在越-英语音翻译任务上提升了1.97个BLEU值。”

Chinese Named Entity Recognition (NER) has continued to attract research attention. However, most existing studies only explore the internal features of the Chinese language but neglect other lingual modal features. Actually, as another modal knowledge of the Chinese language, English contains rich prompts about entities that can potentially be applied to improve the performance of Chinese NER. Therefore, in this study, we explore the bilingual enhancement for Chinese NER and propose a unified bilingual interaction module called the Adapted Cross-Transformers with Global Sparse Attention (ACT-S) to capture the interaction of bilingual information. We utilize a model built upon several different ACT-Ss to integrate the rich English information into the Chinese representation. Moreover, our model can learn the interaction of information between bilinguals (inter-features) and the dependency information within Chinese (intra-features). Compared with existing Chinese NER methods, our proposed model can better handle entities with complex structures. The English text that enhances the model is automatically generated by machine translation, avoiding high labour costs. Experimental results on four well-known benchmark datasets demonstrate the effectiveness and robustness of our proposed model.

Intelligent medical services have attracted great research interests for providing automated medical consultation. However, the lack of corpora becomes a main obstacle to related research, particularly data from real scenarios. In this paper, we construct RealMedDial, a Chinese medical dialogue dataset based on real medical consultation. RealMedDial contains 2,637 medical dialogues and 24,255 utterances obtained from Chinese short-video clips of real medical consultations. We collected and annotated a wide range of meta-data with respect to medical dialogue including doctor profiles, hospital departments, diseases and symptoms for fine-grained analysis on language usage pattern and clinical diagnosis. We evaluate the performance of medical response generation, department routing and doctor recommendation on RealMedDial. Results show that RealMedDial are applicable to a wide range of NLP tasks with respect to medical dialogue.

We demonstrate that knowledge distillation can be used not only to reduce model size, but to simultaneously adapt a contextual language model to a specific domain. We use Multilingual BERT (mBERT; Devlin et al., 2019) as a starting point and follow the knowledge distillation approach of (Sahn et al., 2019) to train a smaller multilingual BERT model that is adapted to the domain at hand. We show that for in-domain tasks, the domain-specific model shows on average 2.3% improvement in F1 score, relative to a model distilled on domain-general data. Whereas much previous work with BERT has fine-tuned the encoder weights during task training, we show that the model improvements from distillation on in-domain data persist even when the encoder weights are frozen during task training, allowing a single encoder to support classifiers for multiple tasks and languages.

Implicit discourse relation recognition is a challenging task due to the lack of connectives as strong linguistic clues. Previous methods primarily encode two arguments separately or extract the specific interaction patterns for the task, which have not fully exploited the annotated relation signal. Therefore, we propose a novel TransS-driven joint learning architecture to address the issues. Specifically, based on the multi-level encoder, we 1) translate discourse relations in low-dimensional embedding space (called TransS), which could mine the latent geometric structure information of argument-relation instances; 2) further exploit the semantic features of arguments to assist discourse understanding; 3) jointly learn 1) and 2) to mutually reinforce each other to obtain the better argument representations, so as to improve the performance of the task. Extensive experimental results on the Penn Discourse TreeBank (PDTB) show that our model achieves competitive results against several state-of-the-art systems.

Homographic puns have a long history in human writing, widely used in written and spoken literature, which usually occur in a certain syntactic or stylistic structure. How to recognize homographic puns is an important research. However, homographic pun recognition does not solve very well in existing work. In this work, we first use WordNet to understand and expand word embedding for settling the polysemy of homographic puns, and then propose a WordNet-Encoded Collocation-Attention network model (WECA) which combined with the context weights for recognizing the puns. Our experiments on the SemEval2017 Task7 and Pun of the Day demonstrate that the proposed model is able to distinguish between homographic pun and non-homographic pun texts. We show the effectiveness of the model to present the capability of choosing qualitatively informative words. The results show that our model achieves the state-of-the-art performance on homographic puns recognition.

In this paper, we study the problem of geometric reasoning (a form of visual reasoning) in the context of question-answering. We introduce Dynamic Spatial Memory Network (DSMN), a new deep network architecture that specializes in answering questions that admit latent visual representations, and learns to generate and reason over such representations. Further, we propose two synthetic benchmarks, FloorPlanQA and ShapeIntersection, to evaluate the geometric reasoning capability of QA systems. Experimental results validate the effectiveness of our proposed DSMN for visual thinking tasks.

This paper introduces Team Alibaba’s systems participating IJCNLP 2017 shared task No. 2 Dimensional Sentiment Analysis for Chinese Phrases (DSAP). The systems mainly utilize a multi-layer neural networks, with multiple features input such as word embedding, part-of-speech-tagging (POST), word clustering, prefix type, character embedding, cross sentiment input, and AdaBoost method for model training. For word level task our best run achieved MAE 0.545 (ranked 2nd), PCC 0.892 (ranked 2nd) in valence prediction and MAE 0.857 (ranked 1st), PCC 0.678 (ranked 2nd) in arousal prediction. For average performance of word and phrase task we achieved MAE 0.5355 (ranked 3rd), PCC 0.8965 (ranked 3rd) in valence prediction and MAE 0.661 (ranked 3rd), PCC 0.766 (ranked 2nd) in arousal prediction. In the final our submitted system achieved 2nd in mean rank.