2604.07895

Total: 1

#1 DialBGM: A Benchmark for Background Music Recommendation from Everyday Multi-Turn Dialogues [PDF] [Copy] [Kimi] [REL]

Authors: Joonhyeok Shin, Jaehoon Kang, Yujun Lee, Hannah Lee, Yejin Lee, Yoonji Park, Kyuhong Shim

Selecting an appropriate background music (BGM) that supports natural human conversation is a common production step in media and interactive systems. In this paper, we introduce dialogue-conditioned BGM recommendation, where a model should select non-intrusive, fitting music for a multi-turn conversation that often contains no music descriptors. To study this novel problem, we present DialBGM, a benchmark of 1,200 open-domain daily dialogues, each paired with four candidate music clips and annotated with human preference rankings. Rankings are determined by background suitability criteria, including contextual relevance, non-intrusiveness, and consistency. We evaluate a wide range of open-source and proprietary models, including audio-language models and multimodal LLMs, and show that current models fall far short of human judgments; no model exceeds 35% Hit@1 when selecting the top-ranked clip. DialBGM provides a standardized benchmark for developing discourse-aware methods for BGM selection and for evaluating both retrieval-based and generative models.

Subject: Artificial Intelligence

Publish: 2026-04-09 07:06:26 UTC