AI Paper Atlas

Research map / Image generation / Text-to-image diffusion

CRAFT-LoRA: Content-Style Personalization via Rank-Constrained Adaptation and Training-Free Fusion

Yu Li, Yujun Cai, Chi Zhang

CVPR 2026 · Image generation › Text-to-image diffusion · approach: Multi-subject personalization · 2 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add time-dependent scaling of LoRA weights in ACFG (ablation)ACFG · ACFG ablationquality score: 0.78 -> 0.86Improved
add low-rank constraint on temporal adjustments (ablation)ACFG · ACFG ablationquality score: 0.82 -> 0.86Improved
anchor the unconditional path to the rank-limited backbone (asymmetric CFG) (ablation)ACFG · ACFG ablationquality score: 0.75 -> 0.86Improved
replace fixed weights with ACFGLoRA fusion · complex composition pairingcomposite quality score: 0.65 -> 0.80Improved
replace fixed weights with ACFGLoRA fusion · distant semantics pairingcomposite quality score: 0.68 -> 0.79Improved
replace fixed weights with ACFGLoRA fusion · similar semantics pairingcomposite quality score: 0.81 -> 0.87Improved
add rank-limited trunk fine-tuning plus layer-selective LoRAStandard LoRA · Content-style disentanglementcross-influence Sx: 0.42 -> 0.18Improved
use hierarchical rank schedule (128 to 4) instead of uniform rank 64 (ablation)Rank-limited backbone fine-tuning · content-style personalizationContent Sim / Style Sim / cross-influence: Content Similarity by 4.8%, Style Similarity by 5.2%, and reduces cross-influence by 11.3%Improved
add Rank-FT to direct merging (ablation)CRAFT-LoRA · component ablationContent Sim / Style Sim: +0.08 Content Sim, +0.10 Style SimImproved
increase contrastive pairs from 50 to 200 (ablation)Rank-limited backbone fine-tuning · content-style personalizationoverall performance: 50 pairs already achieve 95% of full performance, while increasing to 200 pairs yields less than 2% additional improvemeNo clear effect
combine rank-limited updates with contrastive pairsCRAFT-LoRA backbone · content-style personalizationCLIP content-style similarity: 12% reductionImproved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy