Read the original at HF Daily Papers
Researchers study large language models that continue reasoning despite no-thinking instructions, evaluating six models across three question types using three specific metrics to measure compliance.
Carried by: HF Daily Papers. First seen: .