IGen: Redefining long-term event prediction with iterative generation and dynamic balancing

Yuxin Zhang, Yan Wang, Songlin Zhai, Yongrui Chen, Shenyu Zhang, Yuan Meng, Zhihua Chai, Sheng Bi, Guilin Qi

Abstract: Highlights•Pioneer an iterative framework for enhanced long-term event prediction.•Design dual verification and dynamic sampling to balance factuality and diversity.•Demonstrate IGen’s superior performance via extensive benchmark experiments.