AI & Creative Tools
Adobe's Own Researchers Tested AI Image Editors Against Working Professionals — Most Failed Nearly Half the Time
A new arXiv paper, EditAgent, benchmarks AI image-editing agents against professional judgment rather than automated metrics: a general-purpose editor produced no usable result in 42.7% of 600 tasks, specialized tools recovered about two-thirds of those failures, and professionals preferred parallel-generated variants over iterative edits 78.5% of the time.