Newest papers

Predicting Music Valence and Arousal via Textual and Multimodal In-Context Learning in Large Language Models

Chi-En Dai, Chih-Pin Tan, Yi-Hsuan Yang · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.31202 · cited by 0 works (OpenAlex) · open-access PDF

Portuguese

Enquanto a excitação pode ser previsível de forma confiável a partir de áudio musical, a valência emocional permanece difícil de modelar devido à sua natureza subjetiva e dependente do contexto. Neste trabalho, investigamos o potencial da aprendizagem em contexto (ICL) com grandes modelos de idioma para avançar a valência contínua e a previsão de excitação. Em vez de prever diretamente as pontuações emocionais, nosso framework aproveita demonstrações contextuais de um conjunto de dados auxiliares para refinar as previsões iniciais produzidas por um pequeno modelo de base específico de tarefas. Nós avaliamos sistematicamente uma variante de ICL baseada em texto usando letras e capas de áudio geradas, juntamente com uma variante multimodal que incorpora sinais de áudio brutos. Usando todo o conjunto de dados MER60 estritamente

Machine translation (Meta M2M100); the English original follows.

Abstract

While arousal can be reliably predicted from musical audio, emotional valence remains difficult to model due to its subjective and context-dependent nature. In this work, we investigate the potential of in-context learning (ICL) with large language models to advance continuous valence and arousal prediction. Rather than predicting emotion scores directly, our framework leverages contextual demonstrations from an auxiliary dataset to refine the initial predictions produced by a small, task-specific base model. We systematically evaluate a text-based ICL variant using lyrics and generated audio captions, alongside a multimodal variant that incorporates raw audio signals. Using the entire MER60 dataset strictly as a held-out test set, our experiments demonstrate substantial performance gains, particularly for valence. Text-based ICL improves the valence $R^2$ from 0.110 to 0.427 with 80 demonstration shots, while the multimodal variant achieves 0.346 with only 10 shots. These findings suggest that training-free ICL offers an effective inference-time strategy for current language models to predict continuous music emotion.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Valence (chemistry), Arousal, Cognitive psychology, Computer science, Speech recognition, Psychology

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange