'Like gene editing for neural networks': MIT researchers used GPT-6 Astra to edit Qwen's open weights directly, boosting math performance

In the strongest example of "weight space programming", as the technique is called, Astra’s intervention raised Qwen3.5-0.8B’s MATH-500 score from 7.9% to 48.1% while touching approximately 0.016% of ...
OfficeChai

AI Models Can Understand And Improve Model Weights Through “Weight-Space Programming,” Says MIT Research

A new MIT CSAIL blog post is making the rounds in AI circles this week, and it describes something that sounds almost ...
kucoin

MIT researchers use GPT-6 Astra to enhance Qwen's mathematical ability by 6x

Article by Shenchao TechFlowNo training required, no data needed, no gradient descent.Researchers at MIT fed the weights of ...
GitHub

Constrained MEM vs a neural network for DLS inversion

A neural network vs constrained maximum entropy on a synthetic light-scattering inversion problem: where the network breaks, and why. - heliosight/when-neural-networks-fail ...
news.lavx.hu

LLMs read and edit neural weights without gradients

Astra identifies game policies from raw weight matrices and fixes reasoning loops in Qwen3.5-0.8B by editing 0.016% of parameters, lifting MATH-500 scores from 7.9% to 48.1% without training loops.
GitHub

Logs: LLMs can understand and write neural network weights

Complete logs of the experiments in the blog post 'LLMs can understand and write neural network weights' - RikiyaT/lm_understand-logs ...
eu.36kr

MIT allowed GPT-6 Astra to directly modify Qwen's weights, and its math scores have soared.

MIT has unveiled an innovative, groundbreaking paradigm dubbed "Weight Space Programming", representing a transformative new ...