Chia-Yun Lee

Author directory

2026

This paper describes our system for SemEval-2026 Task 3, which focuses on predicting continuous valence and arousal scores. The task poses significant challenges due to variations in data scale and pragmatic ambiguities across languages. To address these disparities, we propose an Adaptive Dual-Track Framework that dynamically selects modeling strategies based on task characteristics. For semantically stable tasks, we apply a robust single baseline optimized with layer-wise learning rate decay (LLRD) to ensure stability. For high-ambiguity scenarios such as the Environmental Protection domain, we adopt a heterogeneous ensemble strategy to mitigate prediction variance. Experimental results demonstrate that our system consistently outperforms the initial standard baseline across all subtasks. Furthermore, our lightweight approach exhibits remarkable parameter efficiency, achieving highly competitive performance against newly introduced large language model (LLM) baselines. Additionally, ablation studies reveal that under regression settings, conventional regularization techniques, cross-lingual data transfer, and homogeneous ensemble learning can lead to negative transfer, confirming the necessity of strategically diverging approaches tailored to linguistic characteristics.
Procrastination is driven by emotional self-regulation failures and cognitive distortions that narratives often mask. We introduce ProcrastiType-ZH, to our knowledge the first Traditional Chinese multi-dimensional procrastination dataset, with 1,000 LLM-generated narratives and human consensus labels for four motives on a 1-5 severity scale (effective range 1-4). We propose ProcrastiCoT, a CBT-informed Chain-of-Thought framework, and inject its LLM-generated reasoning into BERT-size encoders either by concatenation or through a Dual-Encoder Cross-Attention architecture that gives the reasoning its own context window. Across backbones, reasoning injection gives the best single-run QWK on Avoidant, Decisional, and Irrationality. On our primary CKIP-BERT backbone, the Dual-Encoder scores above its pure-encoder counterpart on every dimension in each of four paired training runs, significantly (uncorrected) on Arousal, Decisional, and Irrationality but not Avoidant; on Decisional its mean (0.316) only matches a lexical baseline. A Qwen2.5-7B model fine-tuned on the same labels, with about 70 times more parameters, is strongest on Arousal (QWK 0.709) but below the Dual-Encoder on Irrationality and Avoidant in all four runs. Adding the teacher’s reasoning to its targets lowers QWK on every dimension and pulls its predictions toward the teacher’s ratings, consistent with a conflict between that reasoning and the human labels.

2025

Mental health concerns have garnered increasing attention, highlighting the importance of timely and accurate identification of individual stress states as a critical research domain. This study employs the multimodal StressID dataset to evaluate the contributions of three modalities—physiological signals, video, and audio—in stress recognition tasks. A set of machine learning models, including Random Forests (RF), Support Vector Machines (SVM), Multi-Layer Perceptrons (MLP), and K-Nearest Neighbors (KNN), were trained and tested with optimized parameters for each modality. In addition, the effectiveness of different multimodal fusion strategies was systematically examined. The unimodal experiments revealed that the physiological modality achieved the highest performance in the binary stress classification task (F1-score = 0.751), whereas the audio modality outperformed the others in the three-class classification task (F1-score = 0.625). In the multimodal setting, feature-level fusion yielded stable improvements in the binary classification task, while decision-level fusion achieved superior performance in the three-class classification task (F1-score = 0.65). These findings demonstrate that multimodal integration can substantially enhance the accuracy of stress recognition. Future research directions include incorporating temporal modeling and addressing data imbalance to further improve the robustness and applicability of stress recognition systems.