Astra can read and write weights to improve Qwen

Re:Astra can read and write weights to improve Qwen by ACCount39 on Monday November 02, 2026 @07:57PM (#50346563)

> That weights have some universal properties? E.g., “breakout game weights” always look the same regardless of training and model?

That's exactly what Platonic Representation Hypothesis would predict.

There also have been plenty of studies showing that various neural networks, including LLMs, develop implicit world models that reflect the geometry/structure of a task they're trained on. A trained "chess engine LLM" would have a "chessboard-shaped" world model inside it.

So, if you can figure out what the geometry of a task is, you can search the weights for correlates of that geometry. Then you can try to improve those correlates so that they track the task better.

Seems like bleeding edge LLMs might already be superhuman at "figure out geometry from looking at weights".

[ Parent | on HN ]

Re:Astra can read and write weights to improve Qwen by Vaushite on Monday November 02, 2026 @08:03PM (#50346623)

Even with PRH you can apply orthogonal matrices to intermediate vectors to generate different weights with similar effects.

Even if a slight rotation passing through a nonlinearity isn't good enough, you can just reorder the vector components which makes literally no difference.

[ Parent | on HN ]

Re:Astra can read and write weights to improve Qwen by ACCount39 on Monday November 02, 2026 @08:20PM (#50346766)

Representational similarity analysis is concerned with representation structure independent from the coordinate system it sits in. A map rotated 45 degrees is still the same map.

A "reorder the vector components" transform preserves the similarity perfectly. So does a transform that maintains orthogonal equivalence.

[ Parent | on HN ]