Rethinking Data Selection at Scale: Random Selection is Almost All You Need

Open in new window