MIT 研究显示 GPT-6 Astra 可直接改权重提升 Qwen3.5-0.8B 数学能力
MASSIVE. MIT researchers showed that GPT-6 Astra can improve another AI model by changing its weights directly, without training.
MIT 研究者让 GPT-6 Astra 在 Python notebook 中直接编辑 Qwen3.5-0.8B 的权重，在 MATH-500 上把准确率从 7.9% 提升到 48.1%，全程不训练。Astra 获得权重和 50 次练习数学题测试预算后，发现模型反复重复直到触及长度上限，将这一习惯定位到 2 层写入的单个内部信号，并从权重中减去该信号。这次编辑只改动了 0.016% 的权重，作者将这一方法称为 Weight-Space Programming。
查看原帖我们让 GPT-6 Astra 编辑 Qwen3.5-0.8B 的权重以提升数学能力。 它仅直接改动了 0.016% 的权重，就修复了 0.8B 的推理循环，将数学能力最高提升 6 倍，甚至将指令遵循能力提升 3 倍，全程无需梯度或训练。 我们将此称为权重空间编程（Weight-Space Programming）：前沿大语言模型如今能够原生读写神经网络权重！🧵 博客：https://t.co/bOWCcFhCVq
查看原文
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
正文
MASSIVE. MIT researchers showed that GPT-6 Astra can improve another AI model by changing its weights directly, without training.
Here, GPT-6 Astra raised Qwen3.5-0.8B from 7.9% to 48.1% on MATH-500 by editing 0.016% of its weights, without training.
Astra got Qwen’s weights in a Python notebook, plus a budget of 50 test runs on practice math problems. It saw the model kept repeating itself until it hit its length limit, traced that habit to a single internal signal written by 2 of its layers, and subtracted that signal out of the weights.
With the loop gone, the model could finish answers it was already capable of solving.
That single targeted edit changed 0.016% of the weights and lifted math accuracy from 7.9% to 48.1%.