ScholarEvolve Turns Research Papers Into Code to Improve AI Agents Without Retraining
🤖 AI-generated content — The title and summary were produced automatically by artificial intelligence, without human editorial review.
University of California, Santa Barbara, and Microsoft researchers built ScholarEvolve to convert published research into agent-harness upgrades. It raised Qwen3.5-27B’s AppWorld Normal task completion from 69.0% to 81.4% without changing the model’s weights, closing about 74% of the gap with GPT-5.4.
Continue in the app — vote & join in ➔Source: VentureBeat · via ahirlevel.hu