▲ 2 pointsWeak-to-strong generalization (GPT-4 Turbo was Pareto-SOTA)forourposterity.comby ronfriedhaber·10d ago·0 comments·view on hn ↗