1 article found
Emerging benchmarks show Jev beating traditional ML pipelines on zero-shot tasks across the board. But is it really replacing sklearn, or just making us rethink what we measure?