Salesforce AI Research framework lifts agent pass rate from 43.5% to 93.0%

Verifying
  • Salesforce AI Research published DarwinX, a framework for evolving AI-agent harnesses without changing model weights.
  • 43.5% to 93.0% was the WebArena-Infinity audit-clean pass@1 improvement across 1,260 tasks.
  • The July 31, 2026 arXiv paper was authored by 12 researchers led by Yifan Zhang, Yutong Dai and Juntao Tan.
The information on this website is generated using AI and we cannot guarantee its accuracy. Please use it as reference information only.