The SemEval-2025 Task 11, Bridging the Gap in Text-Based Emotion Detection, introduces an emotion recognition challenge spanning over 28 languages. This competition encourages researchers to explore more advanced approaches to address the challenges posed by the diversity of emotional expressions and background variations. It features two tracks: multi-label classification (Track A) and emotion intensity prediction (Track B), covering six emotion categories: anger, fear, joy, sadness, surprise, and disgust.In our work, we systematically explore the benefits of two contrastive learning approaches: sample-based (Contrastive Reasoning Calibration) and generation-based (DPO, SimPO) contrastive learning. The sample-based contrastive approach trains the model by comparing two samples to generate more reliable predictions. The generation-based contrastive approach trains the model to differentiate between correct and incorrect generations, refining its prediction. All models are fine-tuned from LLaMa3-Instruct-8B. Our system achieves 12th place in Track A and 7th place in Track B for English, while ranking among the top-tier performing systems for other languages.

As the Large Language Model (LLM) gains widespread adoption, increasing attention has been given to the challenge of making LLM forget non-compliant data memorized during its pre-training. Machine Unlearning focuses on efficiently erasing sensitive information from LLM under limited computational resources. To advance research in this area, SemEval 2025 Task 4: “Unlearning Sensitive Content from Large Language Models” introduces three unlearning datasets and establishes a benchmark by evaluating both forgetting effectiveness and the preservation of standard capabilities. In this work, we propose a more controllable forgetting loss, Effective Unlearning Loss, and explore its integration with various techniques to achieve more efficient and controlled unlearning. Our system ultimately ranked 5th on the competition leaderboard.

Encoder-decoder transformer models have achieved great success on various vision-language (VL) and language tasks, but they suffer from high inference latency. Typically, the decoder takes up most of the latency because of the auto-regressive decoding. To accelerate the inference, we propose an approach of performing Dynamic Early Exit on Decoder (DEED). We build a multi-exit encoder-decoder transformer model which is trained with deep supervision so that each of its decoder layers is capable of generating plausible predictions. In addition, we leverage simple yet practical techniques, including shared generation head and adaptation modules, to keep accuracy when exiting at shallow decoder layers. Based on the multi-exit model, we perform step-level dynamic early exit during inference, where the model may decide to use fewer decoder layers based on its confidence of the current layer at each individual decoding step. Considering different number of decoder layers may be used at different decoding steps, we compute deeper-layer decoder features of previous decoding steps just-in-time, which ensures the features from different decoding steps are semantically aligned. We evaluate our approach with three state-of-the-art encoder-decoder transformer models on various VL and language tasks. We show our approach can reduce overall inference latency by 20%-74% with comparable or even higher accuracy compared to baselines.

This paper presents the Chinchunmei team’s contributions to the WASSA2024 Shared-Task 1: Empathy Detection and Emotion Classification. We participated in Tracks 1, 2, and 3 to predict empathetic scores based on dialogue, article, and essay content. We choose Llama3-8b-instruct as our base model. We developed three supervised fine-tuning schemes: standard prediction, role-play, and contrastive prediction, along with an innovative scoring calibration method called Contrastive Reasoning Calibration during inference. Pearson Correlation was used as the evaluation metric across all tracks. For Track 1, we achieved 0.43 on the devset and 0.17 on the testset. For Track 2 emotion, empathy, and polarity labels, we obtained 0.64, 0.66, and 0.79 on the devset and 0.61, 0.68, and 0.58 on the testset. For Track 3 empathy and distress labels, we got 0.64 and 0.56 on the devset and 0.33 and 0.35 on the testset.

This paper presents our participation to the WASSA 2024 Shared Task on Empathy Detection and Emotion Classification and Personality Detection in Interactions. We focus on Track 2: Empathy and Emotion Prediction in Conversations Turns (CONV-turn), which consists of predicting the perceived empathy, emotion polarity and emotion intensity at turn level in a conversation. In the method, we conduct BERT and DeBERTa based finetuning, implement the CombinedLoss which consists of a structured contrastive loss and Pearson loss, adopt adversarial training using Fast Gradient Method (FGM). This method achieved Pearson correlation of 0.581 for Emotion,0.644 for Emotional Polarity and 0.544 for Empathy on the test set, with the average value of 0.590 which ranked 4th among all teams. After submission to WASSA 2024 competition, we further introduced the segmented mix-up for data augmentation, boosting for ensemble and regression experiments, which yield even better results: 0.6521 for Emotion, 0.7376 for EmotionalPolarity, 0.6326 for Empathy in Pearson correlation on the development set. The implementation and fine-tuned models are publicly-available at https://github.com/hyy-33/hyy33-WASSA-2024-Track-2.

In this paper, we describe an ongoing effort in collecting and annotating a multilingual speech database of natural stress emotion from university students. The goal is to detect natural stress emotions and study the stress expression differences in different languages, which may help psychologists in the future. We designed a common questionnaire of stress-inducing and non-stress-inducing questions in English, Mandarin and Cantonese and collected a first ever, multilingual corpus of natural stress emotion. All of the students are native speakers of the corresponding language. We asked native language speakers to annotate recordings according to the participants' self-label states and obtained a very good kappa inter labeler agreement. We carried out human perception tests where listeners who do not understand Chinese were asked to detect stress emotion from the Mandarin Chinese database. Compared to the annotation labels, these human perceived emotions are of low accuracy, which shows a great necessity for natural stress detection research.