AI Paper Atlas

Research map / Image generation / Editing and style transfer

Spanning the Visual Analogy Space with a Weight Basis of LoRAs

Hila Manor, Rinon Gal, Haggai Maron, Tomer Michaeli, Gal Chechik

ECCV 2026 · Image generation › Editing and style transfer · approach: Instruction-based diffusion editing · 4 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
use basis of 32 rank-4 LoRAs instead of one r=128 LoRALoRWeB · Visual analogy eval (100 tasks, 840 triplets)Edit accuracy (VLM): 5.70 -> 5.94Improved
use basis of 32 rank-4 LoRAs instead of one r=128 LoRALoRWeB · Visual analogy eval (100 tasks, 840 triplets)Pairwise VLM vs LoRA r=128: 57.9Improved
add larger basis N=64 (ablation)LoRWeB · Visual analogy eval (100 tasks, 840 triplets)Edit accuracy (VLM): 5.94 -> 5.48Hurt
add larger rank r=16 (N=32 equivalent) (ablation)LoRWeB · Visual analogy eval (100 tasks, 840 triplets)Edit accuracy (VLM): 5.94 -> 4.92Hurt
use large basis N=32 instead of N=8 at same capacity (ablation)LoRWeB · Visual analogy eval (100 tasks, 840 triplets)Edit accuracy (VLM): 5.49 -> 5.94Improved
use N=32 instead of N=16 (ablation)LoRWeB · Visual analogy eval (100 tasks, 840 triplets)Edit accuracy (VLM): 5.95 -> 5.94No clear effect
use softmax instead of Tanh for mixing coefficients (ablation)LoRWeB · Visual analogy eval (100 tasks, 840 triplets)Edit accuracy (VLM): 4.49 -> 5.94Improved
encode a, a', b separately and concatenate instead of encoding the 2x2 grid (ablation)LoRWeB · Visual analogy eval (100 tasks, 840 triplets)Edit accuracy (VLM): 5.75 -> 5.94Improved
swap CLIP for SigLIP2 encoder (ablation)LoRWeB · Visual analogy eval (100 tasks, 840 triplets)Edit accuracy (VLM): 5.94 -> 5.82No clear effect
add mixture of LoRAs vs single LoRA r=256LoRWeB vs LoRA · Visual analogy eval (100 tasks, 840 triplets)Edit accuracy (VLM): 5.48 -> 5.94Improved
use LoRWeB instead of DIA-KontextLoRWeB · Visual analogy eval (100 tasks, 840 triplets)Edit accuracy (VLM): 3.56 -> 5.94Improved
use LoRWeB instead of PairEdit s=0.6LoRWeB · Visual analogy eval (100 tasks, 840 triplets)LPIPS: 0.24 vs 0.31Hurt

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