EVOLVING DEEPER LLM THINKING

Summary from the TLDR AI newsletter

Contrary to the direct R1 approach, this work from DeepMind uses an evolutionary strategy, guided by a language model, to modify and propose new responses. This leads to strong reasoning ability, specifically in planning tasks that require some exploration.