Astra can read and write weights to improve Qwen

Re:Astra can read and write weights to improve Qwen by AlexErrant on Monday November 02, 2026 @03:15PM (#50343034)

As insane as this result is, next steps are even crazier:

> We are currently building a benchmark and a harness for LLMs’ capability on Weight-Space Programming. We believe this will provide a standardized evaluation to measure and accelerate progress in Weight-Space Programming, turning what currently seems like black magic into a rigorous, reproducible engineering discipline.

Imagine this benchmark lasting as long as Arc-AGI-3 has. This could be the first foothold on hill-climbing mech-interp.

[ Parent | on HN ]

Re:Astra can read and write weights to improve Qwen by reasonableklout on Monday November 02, 2026 @04:26PM (#50344008)

This is the first time it dawned on me that mech interp is a dual-use technology which can be used to improve capabilities. Do you think overall this development will improve human understanding of AI models, or will it lead to faster capability advancement using weight-space changes that no humans read?

[ Parent | on HN ]

Re:Astra can read and write weights to improve Qwen by aesthesia on Monday November 02, 2026 @04:44PM (#50344261)

I think edits made directly to model weights are likely to be more interpretable and controllable than edits made by the current process of shoving a bunch of training data through gradient descent.

[ Parent | on HN ]