{
  "id": 272,
  "url": "https://arxiv.org/abs/2607.03397v1",
  "title": "Efficient bias mitigation in T2I diffusion models using Concept Graphs",
  "summary": "Text-to-Image diffusion models often propagate harmful bias inherited from the training data. Existing bias mitigation techniques typically intervene only at the text encoder or provide inference-time guidance, often leading to generations that collapse into semantically incoherent outputs. To address these limitations, we introduce CO-ALIGN (Concept Ontology Alignment), a novel bias mitigation approach based on concept-graph alignment that operates on the model's internal concept ontology. By a",
  "authors": "Mansi, Avinash Kori, Francesco Leofante",
  "category": "research",
  "topics": "bias-fairness,safety-alignment",
  "orgs": null,
  "regions": null,
  "published_at": "2026-07-03T14:58:56.000Z",
  "fetched_at": "2026-07-14T14:14:24.248Z",
  "source_slug": "arxiv-ethics",
  "source_name": "arXiv",
  "source_homepage": "https://arxiv.org",
  "ethics_ai_record_url": "https://ethics.ai/record/272",
  "original_url": "https://arxiv.org/abs/2607.03397v1",
  "evidence_status": "source-only",
  "attribution": "via ethics.ai"
}