Thinking Machines Lab analyzed why large language models provide different answers to identical questions at a temperature of 0 (always selecting the most probable answer).
The article Thinking Machines wants large language models to give consistent answers every time appeared first on THE DECODER.