AI Paper Atlas

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

IMG: Calibrating Diffusion Models via Implicit Multimodal Guidance

Jiayi Guo, Chuanhao Yan, Xingqian Xu, Yulin Wang, Kai Wang, Gao Huang, Humphrey Shi

ICCV 2025 · Image generation › Text-to-image diffusion · approach: Inference-time T2I improvement · 1 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 IMG re-generation to base diffusion modelSDXL · HPD averageHPS: 28.03 -> 28.56Improved
add IMG on top of the finetuned modelSDXL-DPO · HPD averageHPS: 28.39 -> 28.79Improved
add IMGFLUX · HPD averageHPS: 29.21 -> 29.48Improved
add SLD editing-based alignmentSDXL · HPD averageHPS: 28.03 -> 27.35Hurt
add IMG on top of FLUX for compositional promptsFLUX · T2I-CompBench Colorscore: 0.7852 -> 0.8060Improved
use text embedding instead of MLLM features to guide the aligner (ablation)IMG Implicit Aligner · Pick-a-Pic testPick Score: 22.34 -> 22.38 (vs base model)No clear effect
use original MLLM features rather than no guidance (ablation)IMG Implicit Aligner · Pick-a-Pic testPick Score: 22.22 -> 22.59Improved
finetune the MLLM on Instruct-Pix2Pix editing data (ablation)IMG MLLM · Pick-a-Pic testPick Score: 22.59 -> 22.74Improved
finetune the MLLM on editing instruction task (ablation)LLaVA 1.5-13b · SEED-Data-Edit 300 casesyes rate: 42.3% -> 76.3%Improved
set loss ratio lambda to 1 rather than 0, 0.5 or 2 (ablation)IMG objective · Pick-a-Pic testPick Score: 22.56 -> 22.74 (lambda 0 vs 1); 2 gives 22.50Improved
add DPO component then SPIN component to preference loss (ablation)IMG objective · Pick-a-Pic testPick Score: 22.56 -> 22.65 -> 22.74Improved
swap LLaVA for Qwen2.5-VL-7B as MLLM (ablation)IMG · HPD averageHPS: SDXL 28.56 -> 28.48; FLUX 29.48 -> 29.52No clear effect

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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