New 'Markovian Thinking' technique unlocks a path to million-token AI reasoning

Researchers at Mila have introduced an innovative technique called Markovian Thinking, which significantly enhances the efficiency of large language models in complex reasoning tasks. This breakthrough allows these models to perform lengthy reasoning without the high computational costs that have previously hindered such capabilities. The implementation, known as Delethink, organizes reasoning into manageable chunks, paving the way for more advanced AI applications and potentially transforming how machines understand and process information.