Sparse Rewiring trains 89 neurons in Qwen3-8B to create an owl persona

Max Sloef reports that Sparse Rewiring trains 89 of Qwen3-8B's 442,368 neurons to produce an owl persona that matches the results of full supervised fine-tuning (SFT).

Dongxi NLP repeats the claim and says interpretability researchers should pay attention to it.

Dongxi NLP also asks how Sparse Rewiring differs from so-called weight space programming, a question the report does not address. The source material available here does not include an independent evaluation of the result.