{"slug": "diffu-lora-a-novel-low-rank-adaptation-for-personalized-diffusion-models", "title": "Diffu-LoRA: A Novel Low-Rank Adaptation for Personalized Diffusion Models", "summary": "A new arXiv paper (arXiv:2610.10550v1) introduces Diffu-LoRA, a parameter-efficient method that personalizes text-to-image diffusion models by learning how to allocate low-rank adaptation capacity across layers through gated low-rank components and bilevel optimization. Experiments with Stable Diffusion on subjects from DreamBooth and additional collected datasets showed improved subject fidelity and prompt alignment versus the evaluated fine-tuning baselines, with progressive pruning removing low-gate components to meet a prescribed rank budget while the pretrained backbone stays frozen.", "body_md": "arXiv:2610.10550v1 Announce Type: new \nAbstract: Personalizing text-to-image diffusion models from a few reference images requires preserving subject identity while following prompts that describe new contexts. Full-model fine-tuning is parameter-intensive, whereas low-rank adaptation (LoRA) reduces the number of trainable parameters but leaves open how adaptation capacity should be distributed across layers. We introduce Diffu-LoRA, a parameter-efficient method that learns this allocation through gated low-rank adaptation. Diffu-LoRA inserts trainable low-rank components into the linear layers of Transformer blocks and assigns a learnable gate to each component. Bilevel optimization updates the adaptation weights and gate parameters on separate data splits, while progressive pruning removes components with the lowest gate values to meet a prescribed rank budget. This procedure allocates adaptation capacity nonuniformly across layers while keeping the pretrained backbone frozen. Experiments with Stable Diffusion on subjects from DreamBooth and additional collected datasets show improved overall subject fidelity and prompt alignment relative to the evaluated fine-tuning baselines. Ablation studies examine the contributions of bilevel optimization, progressive pruning, and adapter placement. These results support learned rank allocation as a practical approach to parameter-efficient diffusion model personalization.", "url": "https://wpnews.pro/news/diffu-lora-a-novel-low-rank-adaptation-for-personalized-diffusion-models", "canonical_source": "https://arxiv.org/abs/2610.10550", "published_at": "2026-10-09 04:00:00+00:00", "updated_at": "2026-10-09 04:17:42.307431+00:00", "lang": "en", "topics": ["artificial-intelligence", "generative-ai", "machine-learning", "ai-research"], "entities": ["Diffu-LoRA", "Stable Diffusion", "DreamBooth", "arXiv"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/diffu-lora-a-novel-low-rank-adaptation-for-personalized-diffusion-models", "markdown": "https://wpnews.pro/news/diffu-lora-a-novel-low-rank-adaptation-for-personalized-diffusion-models.md", "text": "https://wpnews.pro/news/diffu-lora-a-novel-low-rank-adaptation-for-personalized-diffusion-models.txt", "jsonld": "https://wpnews.pro/news/diffu-lora-a-novel-low-rank-adaptation-for-personalized-diffusion-models.jsonld"}}