Controlling Reasoning Effort in LLMs

This article by Sebastian Raschka explores the concept of controlling reasoning effort in large language models (LLMs), focusing on how models like OpenAI's GPT-5.6 and DeepSeek-R1 implement multiple reasoning-effort settings. It defines reasoning models, discusses reinforcement learning with verifiable rewards (RLVR) for training, and provides insights into developing models with low-, medium-, and high-effort reasoning modes. The article is aimed at researchers and practitioners in machine learning and AI, offering both standalone explanations and references to deeper resources on reasoning model development.