AI Paper Atlas

Research map / Architectures and efficiency / Pruning and efficient tuning

SSR-Merge: Subspace Signal Routing for Training-Free LoRA Merging in Diffusion Models

Zhengxuan Wei, Yi Dong, Zonghui Li, Xianhui Lin, Xing Liu, Hong Gu, Shaofeng Zhang, Wenbin Li, Qi Fan

ICML 2026 · Architectures and efficiency › Pruning and efficient tuning · approach: LoRA variants and theory · 0 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
replace identity router (Task Arithmetic) with statistics-derived router (ablation)SSR on FLUX.1-dev · DreamBooth 10 subjects, K=9 mergedDINOv2 similarity: 0.5356 -> 0.6713Improved
use SSR instead of DAREFLUX.1-dev LoRA merging · DreamBooth 10 subjects, K=9 mergedCLIP score: 0.7376 -> 0.7850Improved
increase calibration steps from 1 to 20 (ablation)SSR K=9 · DreamBooth 10 subjects, K=9 merged, FLUX.1-devDINOv2 similarity: 0.6713 -> 0.6739No clear effect
use more calibration timesteps (20 vs 1) (ablation)SSR K=9 · FLUX.1-dev K=9 calibrationmerge time: >600s vs 34.26 secondsHurt
increase merged LoRAs from 9 to 21SSR · DreamBooth 10 subjects, FLUX.1-devDINOv2 recovery rate: 90.2% -> 77.0%Hurt
use SSR instead of DARE in multi-concept compositionFLUX.1-dev · Multi-concept composition (K=2-4)Success rate: 0.62 -> 0.91Improved
use SSR instead of DARE for facial editingFLUX.1-dev · FFHQ facial editingArcFace similarity: 0.9471 -> 0.9610Improved
vary calibration prompt template (five templates) (ablation)SSR K=3 · DreamBooth, K=3CLIP score: 0.7915 to 0.8267, var 0.00067No clear effect
add Tikhonov regularization lambda<=0.1 to RegMeanRegMean on FLUX.1 · DreamBooth mergeimage qualityHurt
use SSR instead of TIES on Qwen-ImageQwen-Image · DreamBooth 10 subjects, K=9 mergedDINOv2 similarity: 0.5716 -> 0.7461Improved
use SSR instead of DOGE TA on GLUE8 GLUE tasks · GLUE (8 tasks)average score: 79.9 -> 80.9Improved
use SSR instead of TIES for merge timeFLUX.1-dev K=9 · merge wall-clockseconds: 88.93 -> 34.26Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 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