elvis· @omarsar0 · X·· 20 天前AI 评分65
AI 导读
elvis 转发 MIT CSAIL 团队的研究,该研究发现前沿 LLM 在原生读写神经网络权重上表现远超预期。研究让 GPT-6 Astra 直接修改 Qwen3.5-0.8B 的权重,仅改动 0.016% 的权重就修复了推理循环,数学成绩最高提升 6 倍、指令遵循提升 3 倍,全程无需梯度或训练,团队称之为 Weight-Space Programming。elvis 认为这对递归自我改进(RSI)意义重大,现有 RSI 流程主要靠写代码启动训练任务,而这一能力提供了直接通过权重递归自我改进模型的原生路径。
正文 · 原文
Must-read report! They find that LLMs are surprisingly good at reading and writing neural network weights.
Huge for RSI when you think about it!
Current RSI pipelines focus on writing code to launch training jobs.
This capability could offer a more native path to recursively self-improve models via weights.
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://rikiyat.github.io/blog/weight-space-programming/在 X 查看被引用的帖子
来源:elvis · x.com