Claude 现可协助构建评测并通过 hillclimb 持续优化
中文摘要
Claude现可协助构建评测并利用hillclimb算法持续优化应用性能,开发者可通过官方指南学习相关技能。
English Summary
Claude can now build evaluations and perform hillclimbing to optimize applications, with official guides available for developers to improve their workflows.
Original Excerpt
Claude 现在可以帮助构建评测(evals)并基于评测进行 hillclimb 持续优化。Claude Devs 分享了评测设计指导及 Claude Code 可用来改进应用的相关 skills,详见 https://claude.dev/blog/automating-eval-design-and-hillclimbing/