AnalysisAI ModelsNovember 3, 2026

Qwen 3.8 27B can interpret model weights locally, Willison finds

Read original source →simonwillison.net

Simon Willison reproduced an example from the paper "Sparks of Weight-Space Programming," which found GPT-6 Astra can read and edit raw model weights without training loops. Astra fixed Qwen3.5-0.8B's reasoning loops by editing 0.016% of weights, lifting MATH-500 from 7.9% to 48.1%.

1 source

More stories today

Open the live feed