Adobe Research and UC San Diego propose EditAgent system to improve AI image editing
Current overview (not historical)
Adobe Research and UC San Diego introduced EditAgent, a system designed to make AI image editing more like professional Photoshop use. In blind tests involving 600 editing tasks, the base editor Flux-2-Klein-9B failed 42.7% of the time. EditAgent addresses roughly two-thirds of these failures by calling dedicated tools. The Decoder reports that Adobe is testing how well AI agents can use editing tools like a Photoshop professional would.
Written with AI from the reporting · updated 11/1/2026
Expand the detailed fact route and sources
Story route
A station-by-station route from the first public evidence onward; every station carries its own evidence and uncertainty.
- 1Unclassified fact stationNov 1, 23:40single sourcereport-time onlyrevised
Adobe Research 提出 EditAgent 系统，提升 AI 图像编辑能力
Read original source ↗ The DecoderReplay this moment in the explorerEvidence & revisions · 1 source records
- Occurred (fact side)：
- not recorded
- Published (valid side)：
- Nov 1, 22:56
- First seen (known side)：
- Nov 1, 23:40
Fact stationSingle source · one-sided claimRevisedReport-time onlyCopy may have been revised later; time replay is bounded. Layer labels group reading only, not causation between layers.
Uncertainty marks flag where to check the sources; they are not proven causation.
Report timeline
Explore every side through dated source reports.
- The DecoderSelectedRelated fact: single sourceFact occurrence time unverifiedSource report revisedAdobe tests AI agent that edits images like a Photoshop professional
Report first recorded Nov 1, 23:40
Adobe researchers have built EditAgent, an AI system that edits images more like a human Photoshop professional. In tests of 600 image edits, a general-purpose editor failed nearly 43 percent of the time, while specialized tools solved about two-thirds of those failures. EditAgent plans edits in steps, generates multiple candidate solutions in parallel from the original image, and uses a language model to select the best result based on hard data such as brightness and color changes, rather than retrying the same editor repeatedly.
Read original source ↗
Event activity trend
- Latest comparable
- 3
- Comparable peak
- 10Nov 2, 00:00
- 24h change
- -49%
The chart compares consistently observed participants only. Hover or tap to inspect each hour; use arrow keys to move between observations.
Event discussion
Reader opinions are hosted on GitHub Discussions; posting requires a GitHub account. Comments are not verified facts.