







Replication of 'Sparks of Weight-Space Programming' (Takehi et al., 2026)
Can a Linear Program Do Weight-Space Programming?
Last week, MIT CSAIL published Sparks of Weight-Space Programming. Given the raw weights of a small network, GPT-6 Astra could tell which game it had been …
Alex Krentsel on Twitter / X
i've reproduced the results from the MIT blog post today on weight-space programming. but one additional result i have to add: this is a new, step-change emerging capability that's appeared in just the latest set of models. pic.twitter.com/EhMsQaaOkL— Alex Krentsel (@AlexKrentsel) November 2, 2026
statsleelab/weight_space_programming
Contribute to statsleelab/weight_space_programming development by creating an account on GitHub.
Sparks of Weight-Space Programming: LLMs can natively read & write neural network weights
We find that frontier LLMs are surprisingly good at natively reading and writing neural network weights, using a variety of recent tools and algorithms or even via direct weight editing.

Sparks of Weight-Space Programming: LLMs can natively read & write neural network weights
We find that frontier LLMs are surprisingly good at natively reading and writing neural network weights, using a variety of recent tools and algorithms or even via direct weight editing.

Sparks of Weight-Space Programming: LLMs can natively read & write neural network weights
We find that frontier LLMs are surprisingly good at natively reading and writing neural network weights much like code, sometimes with tools such as a Python kernel and sometimes from the raw weights alone.

Qwen 3.8 27B on my laptop can interpret model weights, too
Yesterday’s hot paper was Sparks of Weight-Space Programming: LLMs can natively read & write neural network weights by Rikiya Takehi, Jayden Teoh, Alex Zhang, and Omar Khattab. The paper found …
Weight-Space-Programmed Qwen3.5-0.8B - a Hugging Face Space by ML-Intern-lab
This app lets you type a math problem and (optionally) set a token limit. It then runs both the original Qwen3.5‑0.8B model and its weight‑edited version, showing the two generated answers side‑by‑...
Weight Space Misfolding
Weight space editing changes a neural network’s knowledge, behaviour and alignment through direct modification of its weights. Model-mediated weight space manipulation, also known as weight space...

Rikiya Takehi on Twitter / X
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:… pic.twitter.com/DwTwzXzSbE— Rikiya Takehi (@rikiyatakehi) November 2, 2026
4. Mathematical science of computation and structured programming

The Next 7000 Programming Languages
Landin’s seminal paper “The next 700 programming languages” considered programming languages prior to 1966 and speculated on the next 700. Half-a-century on, we cast programming languages in a Darwinian ‘tree of life’ and explore languages, their features (genes) and language evolution from the viewpoint of ‘survival of the fittest’.

Let’s explore the universe of programming resources together
by Quincy Larson Let’s explore the universe of programming resources together There are a lot of free programming resources out there. Here’s a list of more than a thousand free programming books. Here’s a list of hundreds of free online university c...

How are we doing on third-party Space implementations? @lenooby09.twinkers.online's got one in Haskell so y'all are slacking
Piano to flute, one weight edit
Can an LLM agent detect and repair a planted shortcut in a genomics CNN without data? A preregistered, blinded test of weight-space programming

Mixed Signals: Weight-Space Programming for EEG and ECG

Weight-space programming for auditing clinical decision networks

Weight-Space Programming for Medical Imaging: Do We Still Need Training Data?
statsleelab/weight_space_programming