Designer-RSI: Evolving Procedural Memory from User Traffic for Agentic Graphic Design
Designer-RSI evolves a procedural memory of natural-language design skills, raising Claude Sonnet 4 execution success from 72.7% to 99.3%.
The paper introduces a continual adaptation framework where a frozen frontier model operates professional design software through more than 230 tools while an external procedural memory of natural-language skills widens with new procedures and deepens by revising existing ones. Over five rounds on 1,406 real user briefs and 1,869 graded trajectories, with no weight updates or human labels, the skill bank grew from 76 to 139 procedures. GenEval2 execution success on Claude-Sonnet-4 rose from 72.7% to 99.3%, and combined widening/deepening reached a 58.5% win rate on 200 held-out briefs.
- Frozen model paired with evolving natural-language procedural memory
- Skill bank grew from 76 to 139 procedures without weight updates
- GenEval2 execution success rose from 72.7% to 99.3% on Claude Sonnet 4
- Combined mechanisms reached 58.5% win rate on held-out briefs
From the original article
1 min read at huggingface.co
Professional graphic design is a long-horizon agentic task in which structured, editable artifacts emerge from many interdependent actions, yet outcomes admit no reliable programmatic oracle.
By Hongyang Du, Lan Yan, Christian Flores · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.