Goal
Goal: a task in TermGrade Environments: 1,004 executable Linux tasks (Harbor dataset). The difficulty: a competent agent will handle the happy path (clean sorted data, no edge cases). But the test below injects many subtle, independent edge cases that each must be individually correct. Half…
The task
The difficulty: a competent agent will handle the happy path (clean sorted data, no edge cases). But the test below injects many subtle, independent edge cases that each must be individually correct. Half measures will cause multiple assertion failures.
Part of ai-and/termgrade-environments.