MER 2025: When Affective Computing Meets Large Language Models

TL;DR

MER2025 advances affective computing by integrating Large Language Models into four tracks that span semi-supervised fixed taxonomy recognition, open-vocabulary fine-grained labeling, descriptive emotion understanding, and emotion-to-personality analysis. It introduces expanded, diverse datasets (MER-SEMI, OV-MERD, MER-Caption+, MDPE) and ground-truth evaluation metrics (including $WAF$, hierarchical grouping, and RMSE) to benchmark LLM-driven and multimodal approaches. Across tracks, AffectGPT and other LLM-based baselines demonstrate strong performance, particularly when trained on larger, higher-quality datasets, while DES emphasizes interpretability through multimodal clues and human evaluation. The protocol and open-source baselines aim to catalyze reproducible progress in MER, enabling broader adoption of LLM-guided, open-vocabulary, and descriptive emotion understanding in real-world HCI contexts.

Abstract

MER2025 is the third year of our MER series of challenges, aiming to bring together researchers in the affective computing community to explore emerging trends and future directions in the field. Previously, MER2023 focused on multi-label learning, noise robustness, and semi-supervised learning, while MER2024 introduced a new track dedicated to open-vocabulary emotion recognition. This year, MER2025 centers on the theme "When Affective Computing Meets Large Language Models (LLMs)".We aim to shift the paradigm from traditional categorical frameworks reliant on predefined emotion taxonomies to LLM-driven generative methods, offering innovative solutions for more accurate and reliable emotion understanding. The challenge features four tracks: MER-SEMI focuses on fixed categorical emotion recognition enhanced by semi-supervised learning; MER-FG explores fine-grained emotions, expanding recognition from basic to nuanced emotional states; MER-DES incorporates multimodal cues (beyond emotion words) into predictions to enhance model interpretability; MER-PR investigates whether emotion prediction results can improve personality recognition performance. For the first three tracks, baseline code is available at MERTools, and datasets can be accessed via Hugging Face. For the last track, the dataset and baseline code are available on GitHub.

Figures (5)

• Figure 1: Demo for four tracks.
• Figure 2: Label number distribution for MER-FG.
• Figure 3: Annotation pipeline for MER-FG.
• Figure 4: Example of two emotion wheels.
• Figure 5: Emotion description length distribution.