EMNLP Findings2025
Model Calibration for Emotion Detection
Abstract
In this paper, we propose a unified approach to model calibration for emotion detection that exploits the complementary strengths of knowledge distillation and the MixUp data augmentation technique to enhance the trustworthiness of emotion detection models. Specifically, we use a MixUp method informed by training dynamics that generates augmented data by interpolating easy-to-learn with ambiguous samples based on their similarity and dissimilarity provided by saliency maps. We use this MixUp method to calibrate the teacher model in the first generation of the knowledge distillation process. To further calibrate the teacher models in each generation, we employ dynamic temperature scaling to update the temperature used for scaling the teacher predictions. We find that calibrating the teachers with our method also improves the calibration of the student models. We test our proposed method both indistribution (ID) and out-of-distribution (OOD). To obtain better OOD performance, we further fine-tune our models with a simple MixUp method that interpolates a small number of OOD samples with ambiguous ID samples.
Ask Lune
Know this paper in minutes, not an afternoon.
Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.
Lineage
- Not All Negatives are Equal: Label-Aware Contrastive Loss for Fine-grained Text Classification cites GoEmotions: A Dataset of Fine-Grained Emotions
- On the Calibration of Pre-trained Language Models using Mixup Guided by Area Under the Margin and Saliency cites Calibrated Language Model Fine-Tuning for In- and Out-of-Distribution Data
Builds on8
- Not All Negatives are Equal: Label-Aware Contrastive Loss for Fine-grained Text ClassificationVarsha Suresh, Desmond C. OngEMNLP 2021 · 71 citations
- Calibrated Language Model Fine-Tuning for In- and Out-of-Distribution DataLingkai Kong, Haoming Jiang, Yuchen Zhuang, Jie Lyu et al.EMNLP 2020 · 47 citations
- On the Calibration of Pre-trained Language Models using Mixup Guided by Area Under the Margin and SaliencySeoyeon Park, Cornelia CarageaACL 2022 · 44 citations
- CancerEmo: A Dataset for Fine-Grained Emotion DetectionTiberiu Sosea, Cornelia CarageaEMNLP 2020 · 37 citations
- GoEmotions: A Dataset of Fine-Grained EmotionsDorottya Demszky, Dana Movshovitz-Attias, Jeongwoo Ko, Alan S. Cowen et al.ACL 2020 · 16 citations
Related papers
- Calibrating Student Models for Emotion-related TasksMahshid Hosseini, Cornelia CarageaEMNLP 2022 · 2 citations
- Rethinking Data Distillation: Do Not Overlook CalibrationDongyao Zhu, Yanbo Fang, Bowen Lei, Yiqun Xie et al.ICCV 2023 · 19 citations
- On the Limitations of Temperature Scaling for Distributions with OverlapsMuthu Chidambaram, Rong GeICLR 2024 · 11 citations
- Tailoring Mixup to Data for CalibrationQuentin Bouniot, Pavlo Mozharovskyi, Florence d'Alché-BucICLR 2025
- Distilling Calibrated Knowledge for Stance DetectionYingjie Li, Cornelia CarageaACL Findings 2023 · 8 citations
