ByteBrief
We're a portrait publication through and through. Turn your phone back and your briefing picks up right where you left it.
(We tried widescreen once. It wasn't us.)
MIT CSAIL researchers published November 2 a study using GPT-6 Astra to edit Qwen's weights, raising its math score from 7.9% to 48.1%. The team touched only 0.016% of weights to achieve this sixfold improvement by correcting repetitive reasoning in the neural network.
Tracked by ByteBrief