Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in Coakley et alK. L. Coakley, T. Snelleman, H. Hoos, and O. E. Gundersen, "The embrace of open science: An analysis of a decade of AI research and 56 800 conference papers," Under Review, 2026..
BNMusic: Blending Environmental Noises into Personalized Music
Authors: Chi Zuo, Martin Møller, Pablo Martínez-Nuevo, Huayang Huang, Yu Wu, Ye Zhu
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Our experiments with comprehensive evaluations on Music Bench, EPIC-SOUNDS, and ESC-50 demonstrate the effectiveness of our framework, highlighting the ability to blend environmental noise with rhythmically aligned, adaptively amplified, and enjoyable music segments, minimizing the noticeability of the noise, thereby improving overall acoustic experiences. |
| Researcher Affiliation | Collaboration | 1School of Computer Science, Wuhan University, China 2Bang & Olufsen A/S, Denmark 3Department of Computer Science, Princeton University, USA 4LIX, École Polytechnique, IP Paris, France |
| Pseudocode | No | The paper describes its methodology in Section 3 'BNMusic framework: Blending Noises into personalized Music' through detailed textual descriptions and mathematical formulas, as well as a pipeline illustration in Figure 2. However, it does not include explicitly labeled pseudocode or algorithm blocks. |
| Open Source Code | Yes | The details of our experiment and implementation are provided in the Sec. 4.1, and the codes would be provided in the supplimentary materials. |
| Open Datasets | Yes | We use EPIC-SOUNDS[10] and ESC-50[31] as noise sources... For music data, we use 5,000 five-second clips from the Music Bench dataset [19]... |
| Dataset Splits | No | The paper describes using 1,000 segments from EPIC-SOUNDS and 300 from ESC-50 datasets for noise sources, and 5,000 music clips from Music Bench. It details how generated music clips are formed (e.g., 1,000 EPIC-SOUNDS clips with five prompts), and mentions "50 samples" for subjective evaluation. However, it does not specify explicit training/test/validation splits (e.g., percentages, sample counts for each split) for the datasets used in their experiments or evaluation protocol. |
| Hardware Specification | Yes | each sample was processed in approximately 5 seconds on an Nvidia 4090 GPU. |
| Software Dependencies | No | The paper mentions applying Pyln-norm [37] for loudness normalization and using the Riffusion model [6] with default settings. However, it does not specify version numbers for any of the software libraries, frameworks, or programming languages used in the implementation. |
| Experiment Setup | Yes | The noise is consistently normalized to -18 d B LUFS in all evaluations. The overall music signal control parameter, α, was set to 0.14 to ensure adaptive amplification remained within a reasonable range. |