A new AI model, developed by researchers at a leading university, demonstrates significant improvements in processing complex, multi-step reasoning tasks. The system uses a novel architecture that breaks down problems into smaller, more manageable sub-tasks, allowing it to outperform existing models on benchmarks requiring logical deduction and planning. Early tests show promising results in fields like scientific research and strategic analysis, where understanding sequential dependencies is crucial. The developers emphasize that the technology is still in the research phase and requires further testing for real-world applications. For the full details, read the complete article at https://technologyreview.com/2024/05/15/new-ai-model-excels-at-complex-reasoning.