We present our approach to the COLING-2025 Regulations Challenge, which evaluates large language models (LLMs) on nine regulatory tasks, such as abbreviation recognition and financial data extraction. To address challenges like domain-specific terminologies and dynamic regulatory contexts, we developed a robust data construction pipeline, integrating proprietary Chinese regulatory data, Fin-GPT datasets, and financial Q&A data. The pipeline applied, but was not limited to, language filtering, semantic screening, and deduplication, resulting in a 30,000-example dataset combining financial regulations and general financial data. Using this dataset, we fine-tuned Llama 3.2-3B-Instruct to create Reg-LLaMA, a specialized model that outperformed baselines on the Regulations Challenge and PIXIU datasets. These results demonstrate the effectiveness of domain-specific data construction in advancing LLMs for regulatory tasks, paving the way for reliable and interpretable AI in regulated industries.

Current work of mental health support primarily utilizes unimodal textual data and often fails to understand and respond to users’ emotional states comprehensively. In this study, we introduce a novel framework that enhances Large Language Model (LLM) performance in mental health dialogue systems by integrating multimodal inputs. Our framework uses visual language models to analyze facial expressions and body movements, then combines these visual elements with dialogue context and counseling strategies. This approach allows LLMs to generate more nuanced and supportive responses. The framework comprises four components: in-context learning via computation of semantic similarity; extraction of facial expression descriptions through visual modality data; integration of external knowledge from a knowledge base; and delivery of strategic guidance through a strategy selection module. Both automatic and human evaluations confirm that our approach outperforms existing models, delivering more empathetic, coherent, and contextually relevant mental health support responses.

Most existing methods of Out-of-Domain (OOD) intent classification rely on extensive auxiliary OOD corpora or specific training paradigms. However, they are underdeveloped in the underlying principle that the models should have differentiated confidence in In- and Out-of-domain intent. In this work, we shed light on the fundamental cause of model overconfidence on OOD and demonstrate that calibrated subnetworks can be uncovered by pruning the overparameterized model. Calibrated confidence provided by the subnetwork can better distinguish In- and Out-of-domain, which can be a benefit for almost all post hoc methods. In addition to bringing fundamental insights, we also extend the Lottery Ticket Hypothesis to open-world scenarios. We conduct extensive experiments on four real-world datasets to demonstrate our approach can establish consistent improvements compared with a suite of competitive baselines.

Mental manipulation, a significant form of abuse in interpersonal conversations, presents a challenge to identify due to its context-dependent and often subtle nature. The detection of manipulative language is essential for protecting potential victims, yet the field of Natural Language Processing (NLP) currently faces a scarcity of resources and research on this topic. Our study addresses this gap by introducing a new dataset, named MentalManip, which consists of 4,000 annotated fictional dialogues. This dataset enables a comprehensive analysis of mental manipulation, pinpointing both the techniques utilized for manipulation and the vulnerabilities targeted in victims. Our research further explores the effectiveness of leading-edge models in recognizing manipulative dialogue and its components through a series of experiments with various configurations. The results demonstrate that these models inadequately identify and categorize manipulative content. Attempts to improve their performance by fine-tuning with existing datasets on mental health and toxicity have not overcome these limitations. We anticipate that MentalManip will stimulate further research, leading to progress in both understanding and mitigating the impact of mental manipulation in conversations.

“Multi-Hop Question Answering (MHQA) is a significant area in question answering, requiringmultiple reasoning components, including document retrieval, supporting sentence prediction,and answer span extraction. In this work, we present the first application of label smoothing tothe MHQA task, aiming to enhance generalization capabilities in MHQA systems while miti-gating overfitting of answer spans and reasoning paths in the training set. We introduce a novellabel smoothing technique, F1 Smoothing, which incorporates uncertainty into the learning pro-cess and is specifically tailored for Machine Reading Comprehension (MRC) tasks. Moreover,we employ a Linear Decay Label Smoothing Algorithm (LDLA) in conjunction with curricu-lum learning to progressively reduce uncertainty throughout the training process. Experimenton the HotpotQA dataset confirms the effectiveness of our approach in improving generaliza-tion and achieving significant improvements, leading to new state-of-the-art performance on theHotpotQA leaderboard.”