When Noise Lowers The Loss: Rethinking Likelihood-Based Evaluation in Music Large Language Models

• URL: http://arxiv.org/abs/2602.02738v1
• Date: Mon, 02 Feb 2026 19:53:12 GMT
• Title: When Noise Lowers The Loss: Rethinking Likelihood-Based Evaluation in Music Large Language Models
• Abstract summary: Music large language models (LLMs) need robust methods of evaluating output quality.<n>We show that models encounter systematically corrupted music, undermining its validity as a standalone quality indicator.<n>Our results highlight a new principle: the shape of the loss curve -- rather than its absolute value -- encodes critical information about the quality of the generated content.
• Score: 4.959996513415697
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: The rise of music large language models (LLMs) demands robust methods of evaluating output quality, especially in distinguishing high-quality compositions from "garbage music". Curiously, we observe that the standard cross-entropy loss -- a core training metric -- often decrease when models encounter systematically corrupted music, undermining its validity as a standalone quality indicator. To investigate this paradox, we introduce noise injection experiment, where controlled noise signal of varying lengths are injected into musical contexts. We hypothesize that a model's loss reacting positively to these perturbations, specifically a sharp increase ("Peak" area) for short injection, can serve as a proxy for its ability to discern musical integrity. Experiments with MusicGen models in the audio waveform domain confirm that Music LLMs respond more strongly to local, texture-level disruptions than to global semantic corruption. Beyond exposing this bias, our results highlight a new principle: the shape of the loss curve -- rather than its absolute value -- encodes critical information about the quality of the generated content (i.e., model behavior). We envision this profile-based evaluation as a label-free, model-intrinsic framework for assessing musical quality -- opening the door to more principled training objectives and sharper benchmarks.

Related papers

• Lost in the Noise: How Reasoning Models Fail with Contextual Distractors [57.31788955167306]
  Recent advances in reasoning models and agentic AI systems have led to an increased reliance on diverse external information.<n>We introduce NoisyBench, a comprehensive benchmark that systematically evaluates model robustness across 11 datasets in RAG, reasoning, alignment, and tool-use tasks.<n>Our evaluation reveals a catastrophic performance drop of up to 80% in state-of-the-art models when faced with contextual distractors.
  arXiv  Detail & Related papers  (2026-01-12T05:43:51Z)
• WoW-Bench: Evaluating Fine-Grained Acoustic Perception in Audio-Language Models via Marine Mammal Vocalizations [67.6147632074449]
  We introduce the World-of-Whale benchmark (WoW-Bench) to evaluate low-level auditory perception and cognition using marine mammal vocalizations.<n>WoW-Bench is composed of a Perception benchmark for categorizing novel sounds and a Cognition benchmark, inspired by Bloom's taxonomy, to assess the abilities to remember, understand, apply, and analyze sound events.<n>Experiments with state-of-the-art LALMs show performance far below human levels, indicating a need for stronger auditory grounding in LALMs.
  arXiv  Detail & Related papers  (2025-08-28T16:29:46Z)
• Scaling Self-Supervised Representation Learning for Symbolic Piano Performance [52.661197827466886]
  We study the capabilities of generative autoregressive transformer models trained on large amounts of symbolic solo-piano transcriptions.<n>We use a comparatively smaller, high-quality subset to finetune models to produce musical continuations, perform symbolic classification tasks, and produce general-purpose contrastive MIDI embeddings.
  arXiv  Detail & Related papers  (2025-06-30T14:00:14Z)
• Machine Unlearning for Robust DNNs: Attribution-Guided Partitioning and Neuron Pruning in Noisy Environments [5.8166742412657895]
  Deep neural networks (DNNs) have achieved remarkable success across diverse domains, but their performance can be severely degraded by noisy or corrupted training data.<n>We propose a novel framework that integrates attribution-guided data partitioning, discriminative neuron pruning, and targeted fine-tuning to mitigate the impact of noisy samples.<n>Our framework achieves approximately a 10% absolute accuracy improvement over standard retraining on CIFAR-10 with injected label noise.
  arXiv  Detail & Related papers  (2025-06-13T09:37:11Z)
• Impact of Noisy Supervision in Foundation Model Learning [91.56591923244943]
  This paper is the first work to comprehensively understand and analyze the nature of noise in pre-training datasets.<n>We propose a tuning method (NMTune) to affine the feature space to mitigate the malignant effect of noise and improve generalization.
  arXiv  Detail & Related papers  (2024-03-11T16:22:41Z)
• High Perceptual Quality Image Denoising with a Posterior Sampling CGAN [31.42883613312055]
  We propose a new approach to image denoising using conditional generative adversarial networks (CGANs) Our goal is to achieve high perceptual quality with acceptable distortion. We showcase our proposed method with a novel denoiser architecture that achieves the reformed denoising goal and produces vivid and diverse outcomes in immoderate noise levels.
  arXiv  Detail & Related papers  (2021-03-06T20:18:45Z)
• Music FaderNets: Controllable Music Generation Based On High-Level Features via Low-Level Feature Modelling [5.88864611435337]
  We present a framework that can learn high-level feature representations with a limited amount of data. We refer to our proposed framework as Music FaderNets, which is inspired by the fact that low-level attributes can be continuously manipulated. We demonstrate that the model successfully learns the intrinsic relationship between arousal and its corresponding low-level attributes.
  arXiv  Detail & Related papers  (2020-07-29T16:01:45Z)
• Learning perturbation sets for robust machine learning [97.6757418136662]
  We use a conditional generator that defines the perturbation set over a constrained region of the latent space. We measure the quality of our learned perturbation sets both quantitatively and qualitatively. We leverage our learned perturbation sets to train models which are empirically and certifiably robust to adversarial image corruptions and adversarial lighting variations.
  arXiv  Detail & Related papers  (2020-07-16T16:39:54Z)
• Perceiving Music Quality with GANs [0.0]
  We propose training a GAN on a large music library, and using its discriminator as a no-reference quality assessment measure of the perceived quality of music. In a listening test with 448 human subjects, where participants rated professionally produced music tracks degraded with different levels and types of signal degradations, we establish a dataset of human rated material. By using the human rated dataset we show that the discriminator score correlates significantly with the subjective ratings, suggesting that the proposed method can be used to create a no-reference musical audio quality assessment measure.
  arXiv  Detail & Related papers  (2020-06-11T09:45:54Z)
• Audio Impairment Recognition Using a Correlation-Based Feature Representation [85.08880949780894]
  We propose a new representation of hand-crafted features that is based on the correlation of feature pairs. We show superior performance in terms of compact feature dimensionality and improved computational speed in the test stage.
  arXiv  Detail & Related papers  (2020-03-22T13:34:37Z)

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.