{"ID":5551729,"CreatedAt":"2026-07-02T01:54:51.863792489Z","UpdatedAt":"2026-07-04T11:27:09.201833898Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2607.00817","arxiv_id":"2607.00817","title":"Training-Free Debiasing of Diffusion Models via CLIP-Guided Denoising Optimization","abstract":"Text-to-image diffusion models achieve impressive visual quality, yet demographic bias remains a challenge, as neutral prompts consistently produce stereotypical representations across gender and race. Existing approaches remain limited by costly retraining or by inference-time interventions that often degrade image quality and semantic alignment. We propose Text Embedding Steering (TES), a training-free framework that mitigates demographic bias by directly optimizing conditional text embeddings during the diffusion process. We show that a two-stage strategy - early-stage global alignment followed by iterative denoising-time refinement with CLIP-based feedback - enables stable and controllable attribute steering without modifying model parameters. Extensive experiments on Stable Diffusion demonstrate that TES outperforms existing training-free baselines in fairness while maintaining competitive image quality. These results highlight that inference-time text embedding optimization is a practical and scalable solution for fairness-aware generation in diffusion models.","short_abstract":"Text-to-image diffusion models achieve impressive visual quality, yet demographic bias remains a challenge, as neutral prompts consistently produce stereotypical representations across gender and race. Existing approaches remain limited by costly retraining or by inference-time interventions that often degrade image qu...","url_abs":"https://arxiv.org/abs/2607.00817","url_pdf":"https://arxiv.org/pdf/2607.00817v1","authors":"[\"Dain Kim\",\"Jinseo Kim\",\"Sungyong Baik\"]","published":"2026-07-01T11:41:40Z","proceeding":"cs.CV","tasks":"[\"cs.CV\"]","methods":"[\"Diffusion Model\"]","has_code":false}