AnalysisAI ModelsNovember 3, 2026

Blog rebuts claim Astra edits Qwen weights by reading raw floats

Read original source →honnibal.dev

The critique says only experiments 1, 2 and 4 used raw weights in context, on networks of 17K–172K weights; every real-language-model test used Python. The headline Qwen3.5-0.8B result also ran the model on up to 50 batches of practice problems.

1 source

More stories today

Open the live feed