AI Paper Atlas

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

T-LoRA: Single Image Diffusion Model Customization Without Overfitting

Vera Soboleva, Aibek Alanov, Andrey Kuznetsov, Konstantin Sobolev

AAAI 2026 · Image generation › Text-to-image diffusion · approach: Text-driven synthesis and semantic guidance · 8 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 timestep-dependent rank masking (Vanilla T-LoRA) to LoRA, rank 64 (ablation)LoRA · Single-image (rank 64)Text similarity (TS): 0.232 -> 0.240Improved
add Ortho-LoRA orthogonal init to Vanilla T-LoRA, rank 64 (ablation)T-LoRA · Single-image (rank 64)Text similarity (TS): 0.240 -> 0.256Improved
add Ortho-LoRA init to Vanilla T-LoRA, rank 4 (ablation)T-LoRA · Single-image (rank 4)Text similarity (TS): 0.259 -> 0.255No clear effect
T-LoRA vs LoRA text similarity, rank 64LoRA · Single-image (rank 64)Text similarity (TS): 0.232 -> 0.256Improved
T-LoRA vs LoRA image similarity, rank 64LoRA · Single-image (rank 64)Image similarity (IS): 0.901 -> 0.900No clear effect
fine-tune only at noisiest timesteps [800,1000] (ablation)T-LoRA · SD-XL LoRAqualitative overfittingHurt
exclude noisiest timesteps from fine-tuning (ablation)T-LoRA · SD-XL LoRAconcept fidelityHurt
lower rmin from 50% to 25% of rank (ablation)T-LoRA · T-LoRA rank 64concept fidelityMixed
initialize with last SVD components of random matrix instead of principal components of W (ablation)T-LoRA · Ortho-LoRATS / ISImproved
T-LoRA vs LoRA with 1, 2, 3 imagesLoRA · Multi-image customizationText similarity (TS): 0.232 -> 0.256 (1 image)Improved
T-LoRA vs LoRA in human concept preservationLoRA · User studyConcept preservation (%): 39.3 vs 60.7Hurt
T-LoRA vs SVDiff on text similaritySVDiff · Single-image customizationText similarity (TS): 0.256 vs 0.295Hurt

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