Four new models of Qwen have been released

Qwen has released four new models, namely WorldPM – 72B, WorldPM – 72B – HelpSteer2, WorldPM – 72B – RLHFLow, and WorldPM – 72B – UltraFeedback. These models are mainly used for preference modeling, that is, scoring the responses of other models, thereby providing effective support in supervised learning. The official pointed out that compared with training from scratch, further training using these pre – trained models can achieve better results. This provides new possibilities for developing more efficient machine learning systems.