MIT CSAIL proposes Weight-Space Programming: GPT-6 Astra directly rewrites the weights of Qwen3.5-0.8B
English text is being prepared.
Rikiya Takehi, a PhD student at MIT CSAIL, and his team propose "Weight-Space Programming," which lets GPT-6 Astra directly read and write neural network weights without training or computing gradients. In the experiment, only about 0.016% of the weights of Qwen3.5-0.8B were modified, and MATH-500 rose from 7.9 to 48.1 while GSM8K rose from 13.8 to 67.1; according to the authors, this fixes the 0.8B model's reasoning loops, improving math ability by up to 6x and instruction following by 3x.
Why read this
SelectedAI score 81
The method involves no training and no gradient computation; by changing only about 0.016% of the weights, it substantially improves the small model's math scores, illustrating the idea that frontier models can directly edit weights.
Topics
View original postWe asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math. It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training. We call this Weight-Space Programming: frontier LLMs can now natively read/write neural network weights! 🧵 Blog: https://t.co/bOWCcFhCVq
View original
We asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math. It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training. We call this Weight-Space Programming: frontier LLMs can now natively read/write neural network weights! 🧵 Blog: https://t.co/bOWCcFhCVq
Body
今日 X 上 AI 最火：MIT CSAIL 的博士生 Rikiya Takehi 和他的团队提出了“权重空间编程”的方法，不训练、不算梯度，让 GPT-6 Astra 直接读写神经网络的权重。实验只修改了 Qwen3.5-0.8B 约 0.016% 的权重，就让 MATH-500 从 7.9 涨到 48.1，GSM8K 从 13.8 涨到 67.1 ... 大模型手搓小模型，指日可待🤟