Gary Marcus on X: Clarifying How Task Horizon Falls with Increasing Accuracy Criterion
Gary Marcus(@GaryMarcus)258 字 (约 2 分钟)
45
The post suggests plotting task completion time horizons at different accuracy thresholds directly on the same chart to improve readability of METR's evaluation graph and emphasizes specifying the task domain as software engineering.
入选理由：建议在图表中直接绘制50%、80%、90%和100%准确率阈值下的任务完成时间范围曲线。
FeaturedTweet#METR#visualization#task horizon#accuracy threshold英文