Sakana AI unveiled AB-MCTS, an open-source algorithm enabling collaborative problem-solving among AI models. This “inference-time scaling” approach enhances existing AI frameworks by introducing a third dimension: model selection.
Traditional systems decide whether to engage in deeper reasoning or broader exploration when confronted with a challenge. AB-MCTS goes further, dynamically choosing the most appropriate AI model for each specific task. This intelligent allocation of resources allows for more efficient and effective solutions to complex problems. By facilitating seamless collaboration between diverse AI models, Sakana AI’s AB-MCTS marks a significant step towards more adaptable and powerful AI systems.
