The dashboard was a wall of green. Word error rate under 5 percent. Intent classification at 94 percent on our eval set. Response appropriateness, graded by a rubric we trusted, sitting comfortably in

Markdown supported · ⌘↵ to post
No responses yet.