Skip to content

Adobe and UC San Diego Introduce EditAgent to Select and Verify Image Edits

1 reports 1 reporting sources Updated 2 hours ago

Understand the event

AI overview

Researchers at Adobe Research and UC San Diego introduced EditAgent, a system that generates candidate image edits using different tools in parallel and selects among them with a language model supplied with verification data. It improved image editors’ results across three benchmarks, but required about ten times as much compute as a single editor. On a benchmark of 600 edits, Qwen3.8-27B selected usable results 81.3 percent of the time, compared with 74.8 percent for EditScore-72B. The approach combines specialized editing tools with a separate selection step, making the added computation a tradeoff for better results.

Generated by AI from reports · Updated 2 hours ago

Read next

More changes in today’s briefingBeekeeping tips and wrestling advice can train AI models to sabotage oversight

Report timeline

Follow the reports to explore different sides of the event.

11/1
  1. Adobe researchers introduce EditAgent for tool selection and verification in image editing

    Researchers at Adobe Research and UC San Diego developed EditAgent, which generates image edits with different tools in parallel and uses a standard language model with verification data to select results. On a 600-edit benchmark, Qwen3.8-27B selected usable results 81.3 percent of the time, versus 74.8 percent for EditScore-72B. EditAgent improved editors across three benchmarks but required about ten times as much compute as a single editor.

Keep up with AI, one email a day

Model releases, price changes and compute developments, with original sources. Free, confirmed by email, and easy to unsubscribe. Read a past issue

Event heat trend

There is not enough continuous observation data to plot a trend yet.