Cheng has been one of the central figures in Chinese IR for over two decades, with the kind of multi-decade publication arc that crosses social-network analysis, web data, and retrieval, and which has lately converged on the generative-retrieval line his group at ICT-CAS publishes. The work differs in emphasis from the equivalent Western literature: more weight on graph and network structure as a retrieval signal, more attention to scaling under the specific data and user-behavior conditions of the Chinese-language web. For ai100 — which audits AI engines that operate in language environments shaped by very different web ecosystems — reading Cheng's line is one of the most useful correctives against treating "IR research" as English-web research wearing a translation hat.

Worth following when
you want to know what retrieval research looks like outside the Anglophone canon, especially the part that focuses on graph and network structure and on non-English-web scale.
Topics
generative retrieval from the ICT-CAS school; graph and network structure as retrieval signal; the long arc of Chinese IR as a research tradition.
Key works
body of work on graph-based and network-aware IR (1990s onward, ICT-CAS Beijing); ongoing generative-retrieval publications from his group; SIGIR and CIKM best-paper output across two decades.