AutoDataBench: Can Agents Write the Data That Feeds the Self-Improvement Loop?
AutoDataBench finds no agent scores above 20 of 100 writing accepted training tasks in 45 minutes.
1 stories, newest first · one card per story
AutoDataBench finds no agent scores above 20 of 100 writing accepted training tasks in 45 minutes.
Today 840 readers · 1.1K views · 843 reads · all time 159K views