AI Paper Atlas

Research map / Image generation / Fast diffusion sampling

DogFit: Domain-guided Fine-tuning for Efficient Transfer Learning of Diffusion Models

Yara Bahram, Mohammadhadi Shateri, Eric Granger

AAAI 2026 · Image generation › Fast diffusion sampling · approach: Guidance for conditional diffusion · 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 domain-aware guidance offset to the fine-tuning loss (DogFit)DiT-XL/2 fine-tuning · Six target domains avgFID: 19.14 -> 12.34Improved
use DogFit instead of CFG sampling guidanceDiT-XL/2 · Six target domains avgFID: 14.46 -> 12.34, 732.28 -> 366.14 TFLOPSImproved
use DogFit instead of MG training-time guidance on SiTSiT-XL/2 · Six target domains avgFID: 10.36 -> 10.91Hurt
use DogFit instead of MG on SiTSiT-XL/2 · Six target domains avgFDDINOV2: 288.70 -> 252.29Improved
add guidance-strength conditioning (DogFit + Control) (ablation)DogFit DiT-XL/2 · Six target domains avgFID: 12.34 -> 12.63No clear effect
add DogFit to unlabeled FFHQ instead of CFG or MGDiT-XL/2 · FFHQFID: 15.94 -> 10.48Improved
use DogFit instead of MG with fewer sampling steps (10) (ablation)DiT-XL/2 · ArtBenchFID: 108.26 -> 88.84Improved
use DogFit instead of MG at 100 sampling steps (ablation)DiT-XL/2 · FoodFID: 7.60 -> 8.09Hurt
combine DogFit with DiffFit PEFT (ablation)DogFit DiT-XL/2 · Food and ArtBenchFID: Food 10.64 -> 11.86; Art 16.32 -> 15.80Mixed
sample guidance w with excessive exposure to high values (large training exposure) (ablation)DogFit controllable guidance · Food and ArtBenchFIDHurt
delay guidance start (late-start) moderately (ablation)DogFit · DiT-XL/2 target domainsFIDImproved
set late-start threshold too high (ablation)DogFit · DiT-XL/2 target domainsPrecisionHurt

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