Luiz Rocha, a zoologist and curator at the California Academy of Sciences, returned to Palau in May 2026 to retrieve autonomous reef monitoring structures (ARMS) that had been deployed in deep reefs since 2016. ^[1] The researchers presented ARMS, a tool-augmented LLM agent that retrieves complete CEDAR metadata templates and dynamically queries authoritative biomedical terminology services at execution time. ^[2] Autonomous reef monitoring structures (ARMS) consist of stacks of PVC plates separated at uniform intervals, designed to act as standardized 'prefab housing' that collects marine life like sponges, corals, and shrimp. ^[3] From 2016 to 2018, Rocha and colleagues deployed ARMS in deep reefs around Palau, Guam, the Marshall Islands, and French Polynesia at depths of 10, 50, and 100 meters. ^[4] Luiz Rocha stated that the deep reefs are 'really, really unknown' and that researchers lack a baseline understanding of the ecosystem before they can ask fundamental questions about it. ^[5] ARMS outperformed the prompt-based approach, increasing precision from 0.56 to 0.93 and recall from 0.51 to 0.85, with improvements across all field categories and assay types. ^[6] The authors concluded that large language models cannot convert legacy metadata to standards-adherent form without knowledge of the relevant standards. ^[7] The study compared ARMS with a prompt-based approach on 839 legacy metadata records from the Human BioMolecular Atlas Program (HuBMAP), using expert-standardized records as the reference standard. ^[8]