Claude Opus 4.7 demonstrates strong substantive accuracy on BigLaw Bench for Harvey, scoring 90.9% at high effort with better reasoning calibration on review tables and noticeably smarter handling of ambiguous document editing tasks.
在法律文档处理中达到90.9%的准确率,特别是在处理模糊文档编辑任务时的智能提升,展示了AI在专业领域的深度应用能力,这种进步将极大扩展AI在法律和合规领域的应用价值。