most existing large language model agent systems face severe limitations in data-intensive settings, including context saturation, cascading error propagation, and high end-to-end latency

The explosion of data-intensive research is challenging publishers to create new solutions to link publications to research data (and vice versa), to facilitate data mining and to manage the dataset as a potential unit of publication. Change continues to be rapid, with new leadership and coordination from the Research Data Alliance (launched 2013): most research funders have introduced or tightened policies requiring deposit and sharing of data; data repositories have grown in number and type (including repositories for “orphan” data); and DataCite was launched to help make research data cited, visible and accessible. Meanwhile publishers have responded by working closely with many of the community-led projects; by developing data deposit and sharing policies for journals, and introducing data citation policies; by linking or incorporating data; by launching some pioneering data journals and services; by the development of data discovery services such as Thomson Reuters’ Data Citation Index (page 138).