Digg
@digg
The internet is a big place. We find the good stuff.
가입 June 2008
578 팔로잉 중 1.1M 팬
MIT CSAIL researchers say GPT-6 Astra edited 0.016% of Qwen3.5-0.8B’s weights without gradients or training, and the model’s MATH-500 score rose from 7.9% to 48.1% in their experiment.
더 보기
