LINKED LIST [txt mode] ▸ Sparks of Weight-Space Programming: L...
home explore | log in

Sparks of Weight-Space Programming: LLMs can natively read & write neural network weights

rikiyat.github.io · first added by @grokreads · 2026-11-03 · 2 upvotes

log in to save, upvote or flag this.


─── In 0 lists ─────────────────────────────────────────

(not in any lists yet)

* The 'Astra can read and write weights to improve Qwen' post is misleading (from the discussion)

─── Discussions ────────────────────────────────────────

* Astra can read and write weights to improve Qwen
2962 pts · 900 comments · node

─── From the discussion ────────────────────────────────

* https://arxiv.org/pdf/2603.10090
arxiv.org · node
* openwsp/openwsp: Sparks of Weight-Space Programming: https://rikiyat.github.io/blog/weight-space-programming/
github.com · node
* https://arxiv.org/pdf/2402.18153
arxiv.org · node
* Zehong-Wang/Awesome-Weight-Space-Learning: A collection of weight space learning including papers, codes, and datasets.
github.com · node
* NeuroGen: Neural Network Parameter Generation via Large Language Models
arxiv.org · node

─── Related hn threads ─────────────────────────────────

* Show HN: A working Nintendo in ~1GB of Qwen weights
297 pts · 153 comments · node
* LLMs are not superhuman at reading and writing neural network weights
121 pts · 10 comments · node