Poor code examples cause LLM misalignment in unrelated domains (via doctor_eval) — discussion

#ai