ScholarEvolve turns papers into agent code without retraining

What the research found

ScholarEvolve turns published papers into agent code and lifts Qwen AppWorld completion from 69% to 81.4% without retraining, though the method's generality is unverified.

Claims and how well the evidence supports them

1. ScholarEvolve raised Qwen AppWorld task completion from 69% to 81.4% 98%
2. The method improves agents without retraining the model 98%
3. Method turns published academic papers into working code (not verified) 17%

Evidence

• venturebeat.com https://venturebeat.com/orchestration/researchers-made-ai-agents-better-by-turning-academic-papers-into-working-code-no-retraining-required

Actionability rated medium.