Signals 4 · free daily AI digest

Mind the RefGAP: Correcting Reference Attention in Diffusion-Based Visual Editing

Paper recorded by Signals 4 on 2026-09-28 in cs.CV. Abstract reproduced from arXiv; link to the original below.

Published 2026-09-28 on arXiv · recorded by Signals 4 on 2026-09-29

Category: cs.CV · 计算机视觉 · first seen 2026-09-29

Abstract

Reference-guided diffusion editors struggle to faithfully reproduce user-provided references. We identify a potential bottleneck in diffusion editors: many methods provide limited reference-attention allocation. For example, in LoomVideo, edit-region queries assign less than 1% of their attention mass to the reference. We introduce RefGAP, a training-free correction that determines logit-offset ma

Read on arXiv →

#9 most recent of 342 cs.CV papers we have recorded · ↑ newer: Lagrangian--Hamiltonian Flows for Video Prediction and Image Generatio · ↓ older: DynaTokens: Teaching Dynamics to Camera-Controlled Video Models at Tes
Cite this page: Mind the RefGAP: Correcting Reference Attention in Diffusion-Based Visual Editing: the #9 most recent of 342 cs.CV papers we have recorded (as of 2026-09-28). Source: Signals 4 (Signals API) — https://data.jiangzhang.ca/signals4/t/papers/mind-the-refgap-correcting-reference-attention-in-diffusion-based-visual-editing.html
Free to quote with attribution to “Signals 4 (Signals API)”. Machine-readable: papers.json
Related: More cs.CV papers · arXiv signals · All papers · Today in AI
Get 4 AI signals a day by email — free.
Subscribe free → See all plans →
Get 4 AI signals a day by email — free
All models · All repos · By company · Daily editions