System-I

Researchers have found that hallucinations in small language models (SLMs) can be useful for solving complex multi-step reasoning problems. A new study proposes a "answer first-reason later" framework…