{
  "abstract": "Real-world audio typically contains concurrent sound events, yet standard self-supervised methods treat each clip as an isolated training unit without explicitly modeling mixture--constituent relationships. We propose Mixture-to-Constituent Latent Alignment (MiCLA), a two-stage self-supervised pre-training framework that uses synthetic log-mel mixing to align mixture representations with their constituent clips. Stage 1 structures single-clip representations via masked self-distillation and contrastive regularization. Stage 2 introduces bidirectional alignment: for a synthetic mixture, MiCLA suppresses the component aligned with one constituent and aligns the residual with the other, using projective residualization on patch tokens and sample-adaptive soft residualization on the global class token. In downstream evaluations, MiCLA achieves 43.2% mean average precision (mAP) on AS-20K and 50.3% mAP on AS-2M under fine-tuning, and achieves the highest mAP among evaluated models on both AudioSet settings under frozen Convex Gated Probing (CGP). Furthermore, a Joint Semantic Retrieval diagnostic shows that MiCLA improves simultaneous retrieval of both constituent classes over target-averaging baselines. These results indicate that directional mixture-to-constituent alignment provides an effective pre-training signal for polyphonic audio representation learning.",
  "abstract_shown": true,
  "affiliations": [],
  "article_type": null,
  "arxiv": "2610.27491",
  "author_total": 2,
  "authors": [
    {
      "name": "Kohei Yamamoto"
    },
    {
      "name": "Kosuke Okusa"
    }
  ],
  "bibtex": "https://research.draconitic.com/paper/ewr2x29wj4q7r.bib",
  "cited_by": 0,
  "doi": "10.48550/arxiv.2610.27491",
  "fulltext_url": null,
  "id": "ewr2x29wj4q7r",
  "issns": [],
  "landing_url": "https://arxiv.org/abs/2610.27491",
  "language": null,
  "licence": "LicenseRef-arXiv-nonexclusive-1.0",
  "licence_label": "arXiv licence",
  "licence_source": "article",
  "licence_url": "http://arxiv.org/licenses/nonexclusive-distrib/1.0/",
  "metadata_licence": "CC0-1.0",
  "native_subjects": [
    [
      "arxiv",
      "cs.LG"
    ],
    [
      "arxiv",
      "cs.SD"
    ],
    [
      "arxiv",
      "stat.ML"
    ],
    [
      "openalex",
      "1711|Music and Audio Processing"
    ]
  ],
  "openalex": "W7223154406",
  "pdf_url": "https://arxiv.org/pdf/2610.27491",
  "pmcid": null,
  "pmid": null,
  "published": "2026-10-20",
  "publisher": null,
  "ris": "https://research.draconitic.com/paper/ewr2x29wj4q7r.ris",
  "sources": [
    {
      "first_seen": "2026-10-30T12:34:13Z",
      "native_id": "2610.27491",
      "source": "arxiv",
      "url": "https://arxiv.org/abs/2610.27491"
    }
  ],
  "subjects": [
    {
      "field": "mathematics",
      "origin": "arxiv:stat.ML",
      "subfield": "statistics"
    }
  ],
  "title": "MiCLA: Mixture-to-Constituent Latent Alignment for Polyphonic Audio Representation Learning",
  "title_text": "MiCLA: Mixture-to-Constituent Latent Alignment for Polyphonic Audio Representation Learning",
  "type": "preprint",
  "updated_at": "2026-10-30T12:34:13Z",
  "url": "https://research.draconitic.com/paper/ewr2x29wj4q7r",
  "venue": "arXiv",
  "versions": []
}