paper.json: A Machine-Readable Convention So LLM Agents Can Actually Parse Academic Papers

arXiv·medium signal

Proposes paper.json, a companion JSON file that travels with the PDF and addresses three recurring failures when LLM agents read papers: sub-claims that cannot be cited at sub-paper granularity, scope overextension beyond what the paper tests, and figure reproduction commands buried in codebases. Each claim gets a citable ID, scope boundaries are explicit, and figure commands are machine-executable. Directly relevant to anyone building research agents or RAG pipelines over academic literature.