4 ms·Inverse Scaling Prize: Second Round Winners2 points by singularity2001 4y agosingularity2001 4y agotasks in which language models perform worse the bigger they get