CMI-RewardBench: Evaluating Music Reward Models with Compositional Multimodal Instruction

• URL: http://arxiv.org/abs/2603.00610v2
• Date: Wed, 04 Mar 2026 18:26:58 GMT
• Title: CMI-RewardBench: Evaluating Music Reward Models with Compositional Multimodal Instruction
• Authors: Yinghao Ma, Haiwen Xia, Hewei Gao, Weixiong Chen, Yuxin Ye, Yuchen Yang, Sungkyun Chang, Mingshuo Ding, Yizhi Li, Ruibin Yuan, Simon Dixon, Emmanouil Benetos,
• Abstract summary: We establish a comprehensive ecosystem for music reward modeling under Compositional Multimodal Instruction (CMI)<n>We first introduce CMI-Pref-Pseudo, a large-scale preference dataset comprising 110k pseudo-labeled samples, and CMI-Pref, a high-quality, human-annotated corpus tailored for fine-grained alignment tasks.<n>To unify the evaluation landscape, we propose CMI-RewardBench, a unified benchmark that evaluates music reward models on heterogeneous samples across musicality, text-music alignment, and compositional instruction alignment.
• Score: 28.66731946887381
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: While music generation models have evolved to handle complex multimodal inputs mixing text, lyrics, and reference audio, evaluation mechanisms have lagged behind. In this paper, we bridge this critical gap by establishing a comprehensive ecosystem for music reward modeling under Compositional Multimodal Instruction (CMI), where the generated music may be conditioned on text descriptions, lyrics, and audio prompts. We first introduce CMI-Pref-Pseudo, a large-scale preference dataset comprising 110k pseudo-labeled samples, and CMI-Pref, a high-quality, human-annotated corpus tailored for fine-grained alignment tasks. To unify the evaluation landscape, we propose CMI-RewardBench, a unified benchmark that evaluates music reward models on heterogeneous samples across musicality, text-music alignment, and compositional instruction alignment. Leveraging these resources, we develop CMI reward models (CMI-RMs), a parameter-efficient reward model family capable of processing heterogeneous inputs. We evaluate their correlation with human judgments scores on musicality and alignment on CMI-Pref along with previous datasets. Further experiments demonstrate that CMI-RM not only correlates strongly with human judgments, but also enables effective inference-time scaling via top-k filtering. The necessary training data, benchmarks, and reward models are publicly available.

Related papers

• Linear Complexity Self-Supervised Learning for Music Understanding with Random Quantizer [0.0]
  This paper focuses on the reduction of a foundation's model size when applied to music information retrieval (MIR) tasks.<n>Our research combines the Branchformer architecture with SummaryMixing, which were first applied in speech recognition, along with a random quantization process.<n>Our results show that our architecture achieves competitive performance when compared with other state-of-the-art models.
  arXiv  Detail & Related papers  (2026-01-14T16:23:31Z)
• SoliReward: Mitigating Susceptibility to Reward Hacking and Annotation Noise in Video Generation Reward Models [53.19726629537694]
  Post-training alignment of video generation models with human preferences is a critical goal.<n>Current data collection paradigms, reliant on in-prompt pairwise annotations, suffer from labeling noise.<n>We propose SoliReward, a systematic framework for video RM training.
  arXiv  Detail & Related papers  (2025-12-17T14:28:23Z)
• Advancing the Foundation Model for Music Understanding [9.210248657997687]
  We introduce a unified foundation model named MuFun for holistic music understanding.<n>Our model features a novel architecture that jointly processes instrumental and lyrical content.<n>We also propose a new benchmark for multi-faceted music understanding called MuCUE.
  arXiv  Detail & Related papers  (2025-08-02T03:33:47Z)
• CMI-Bench: A Comprehensive Benchmark for Evaluating Music Instruction Following [12.638115555721257]
  CMI-Bench is a comprehensive music instruction following benchmark designed to evaluate audio-text LLMs on a diverse set of music information retrieval (MIR) tasks.<n>Unlike previous benchmarks, CMI-Bench adopts standardized evaluation metrics consistent with previous state-of-the-art MIR models.<n>We provide an evaluation toolkit supporting all open-source audio-textual LLMs, including LTU, Qwen-audio, SALMONN, MusiLingo, etc.
  arXiv  Detail & Related papers  (2025-06-14T00:18:44Z)
• Aligning Text-to-Music Evaluation with Human Preferences [63.08368388389259]
  We study the design space of reference-based divergence metrics for evaluating generative acoustic text-to-music (TTM) models.<n>We find that not only is the standard FAD setup inconsistent on both synthetic and human preference data, but that nearly all existing metrics fail to effectively capture desiderata.<n>We propose a new metric, the MAUVE Audio Divergence (MAD), computed on representations from a self-supervised audio embedding model.
  arXiv  Detail & Related papers  (2025-03-20T19:31:04Z)
• Composed Multi-modal Retrieval: A Survey of Approaches and Applications [81.54640206021757]
  Composed Multi-modal Retrieval (CMR) emerges as a pivotal next-generation technology.<n>CMR enables users to query images or videos by integrating a reference visual input with textual modifications.<n>This paper provides a comprehensive survey of CMR, covering its fundamental challenges, technical advancements, and applications.
  arXiv  Detail & Related papers  (2025-03-03T09:18:43Z)
• Machine Learning Framework for Audio-Based Content Evaluation using MFCC, Chroma, Spectral Contrast, and Temporal Feature Engineering [0.0]
  We construct a dataset containing audio samples from music covers on YouTube along with the audio of the original song, and sentiment scores derived from user comments. Our approach involves extensive pre-processing, segmenting audio signals into 30-second windows, and extracting high-dimensional feature representations. We train regression models to predict sentiment scores on a 0-100 scale, achieving root mean square error (RMSE) values of 3.420, 5.482, 2.783, and 4.212, respectively.
  arXiv  Detail & Related papers  (2024-10-31T20:26:26Z)
• Stem-JEPA: A Joint-Embedding Predictive Architecture for Musical Stem Compatibility Estimation [3.8570045844185237]
  We present Stem-JEPA, a novel Joint-Embedding Predictive Architecture (JEPA) trained on a multi-track dataset. Our model comprises two networks: an encoder and a predictor, which are jointly trained to predict the embeddings of compatible stems. We evaluate our model's performance on a retrieval task on the MUSDB18 dataset, testing its ability to find the missing stem from a mix.
  arXiv  Detail & Related papers  (2024-08-05T14:34:40Z)
• Simple and Controllable Music Generation [94.61958781346176]
  MusicGen is a single Language Model (LM) that operates over several streams of compressed discrete music representation, i.e., tokens. Unlike prior work, MusicGen is comprised of a single-stage transformer LM together with efficient token interleaving patterns.
  arXiv  Detail & Related papers  (2023-06-08T15:31:05Z)
• Revisiting the Evaluation of Image Synthesis with GANs [55.72247435112475]
  This study presents an empirical investigation into the evaluation of synthesis performance, with generative adversarial networks (GANs) as a representative of generative models. In particular, we make in-depth analyses of various factors, including how to represent a data point in the representation space, how to calculate a fair distance using selected samples, and how many instances to use from each set.
  arXiv  Detail & Related papers  (2023-04-04T17:54:32Z)
• Score-informed Networks for Music Performance Assessment [64.12728872707446]
  Deep neural network-based methods incorporating score information into MPA models have not yet been investigated. We introduce three different models capable of score-informed performance assessment.
  arXiv  Detail & Related papers  (2020-08-01T07:46:24Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.