#llm-interpretability on Hacker News

2 stories tagged llm-interpretability, newest first.

Screenshot of nathan.rs: Can a Linear Program Do Weight-Space Programming?

Can a Linear Program Do Weight-Space Programming?(nathan.rs)

285 points by emschwartz 2 days ago | 41 comments

Blog post arguing that a model's weight edit fixing Qwen3.5-0.8B thinking loops can be reproduced by solving a linear program, with fewer changed weights.

weight-space-programming linear-programming llm-interpretability qwen

Screenshot of simonwillison.net: Qwen 3.8 27B on my laptop can interpret model weights, too

Qwen 3.8 27B on my laptop can interpret model weights, too (simonwillison.net)

698 points by theanonymousone 10 days ago | 203 comments

Simon Willison blog post testing whether Qwen 3.8 27B running locally can identify which game a neural network was trained on from its raw weights, following a recent paper.

qwen weight-space-programming local-llms llm-interpretability

#llm-interpretability on Hacker News

Can a Linear Program Do Weight-Space Programming?(nathan.rs)

Qwen 3.8 27B on my laptop can interpret model weights, too(simonwillison.net)

Qwen 3.8 27B on my laptop can interpret model weights, too (simonwillison.net)