Large language models (LLMs) can learn complex reasoning tasks without relying on large datasets, according to a new study by researchers at Shanghai Jiao Tong University. Their findings show that ...