Making Image Editing Easier via Adaptive Task Reformulation with Agentic Executions DOI: XXXXXXX.XXXXXXX Conference: Make sure to enter the correct conference title from your rights confirmation email; June 03–05, 2018; Woodstock, NY ISBN: 978-1-4503-XXXX-X/2018/06

Bo Zhao, Kairui Guo, Runnan Du, Haiyang Sun, Pengshan Wang, Huan Yang, Kun Gai, Yixin Cao, Wei Ji · 2026

Read Making Image Editing Easier via Adaptive Task Reformulation with Agentic Executions DOI: XXXXXXX.XXXXXXX Conference: Make sure to enter the correct conference title from your rights confirmation email; June 03–05, 2018; Woodstock, NY ISBN: 978-1-4503-XXXX-X/2018/06 by Bo Zhao, Kairui Guo, Runnan Du, Haiyang Sun, Pengshan Wang, Huan Yang, Kun Gai, Yixin Cao, Wei Ji (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-15917.txt.

Description

Zhao and eight colleagues reframe image editing failures as a task formulation problem, not a lack of model power. Many failures trace to poorly worded tasks, such as tiny targets, vague spatial relations, or thin instructions. Their adaptive reformulation framework turns each image-instruction pair into a sequence of steps chosen and run by a multimodal agent through analysis, routing, reformulation, and feedback. Tests across ImgEdit, PICA, and RePlan show steady gains, largest on hard cases.

Catalog record

CREATORBo Zhao, Kairui Guo, Runnan Du, Haiyang Sun, Pengshan Wang, Huan Yang, Kun Gai, Yixin Cao, Wei Ji
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE56528
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2604.15917

Browse the full library · About the Tower of Babel