Abstract: With the iterative breakthroughs in technological applications of generative artificial intelligence (GAI), an increasing number of researchers have begun utilizing GAI to assist scientific work. This new wave of "technological dividends" and its potential "intelligence gap" warrant close attention. Based on empirical survey data from researchers at multiple Chinese universities, this study employs cutting-edge machine learning algorithms to systematically examine the impacts of GAI on researchers' innovative behaviors, working hours, and their heterogeneous treatment effects. The findings reveal:First, while GAI significantly stimulates researchers' innovative behaviors, it simultaneously leads to a corresponding increase in working hours, indicating that the current "technological dividends" of GAI are primarily manifested in the aspect of innovation incentives.Second, the effects of GAI demonstrate significant heterogeneity among researchers, with research output, age, and gender being critical factors influencing these differential treatment effects. Specifically, male researchers and high-output individuals exhibit more pronounced improvements in innovative behaviors under GAI's influence, while experiencing relatively limited negative impacts on working hours. Further verification confirms that high-performing researchers characterized by superior productivity constitute the main beneficiaries of GAI's "technological dividends" at the current stage.