People are working on using LLMs to regenerate experimental data from the text of journal papers.