AI Paper Atlas

Research map / Image generation / Fast diffusion sampling

Improving Diffusion Models for Class-imbalanced Training Data via Capacity Manipulation

Feng Hong, Jiangchao Yao, Yifei Shen, Dongsheng Li, Ya Zhang, Yanfeng Wang

ICLR 2026 · Image generation › Fast diffusion sampling · approach: Diffusion theory and design space

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 Capacity ManipulationDDPM · Imb. CIFAR-100 (IR=100)FID: 10.163 -> 7.519Improved
add Capacity ManipulationDDPM · Imb. CIFAR-10 (IR=100)FID: 10.697 -> 7.727Improved
add Capacity ManipulationDDPM · Imb. CelebA-HQ (IR=100) maleFID: 16.425 -> 12.788Improved
add Capacity ManipulationDDPM · Imb. iNaturalist 64x64FID: 7.99 -> 6.82Improved
add Capacity ManipulationStable Diffusion LoRA · Imb. ArtBench-10 (IR=100)FID: 27.083 -> 22.776Improved
add consistency loss L_Con (ablation)CM · Imb. CIFAR-100 (IR=100)FID: 8.412 -> 7.519Improved
add diversity loss L_Div (ablation)CM · Imb. CIFAR-100 (IR=100)FID: 8.073 -> 7.519Improved
add reserved minority capacity theta_e at inference (vs theta_g only) (ablation)CM · Imb. CIFAR-100 (IR=100) Few splitFew FID: 29.357 -> 18.729Improved
add reserved minority capacity theta_e at inference (vs theta_g only) (ablation)CM · Imb. CIFAR-100 (IR=100) overallFID: 13.712 -> 7.519Improved
use low-rank decomposition with CBDM objective onlyCBDM · Imb. CIFAR-100 (IR=100)FID: 10.051 -> 10.231No clear effect
fine-tune CBDM with LoRA instead of CMCBDM · Imb. CIFAR-100 (IR=100)FID: 10.051 -> 11.424Hurt
use CM instead of Group-Expert LoRA (MoE-style) (ablation)LoRA adapters · Imb. CIFAR-100FID: 10.06 -> 7.519Improved

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