Source registry

Every research paper page identifies its original source with a stable paper ID and external URL. Missing publication, artifact, license, or acquisition information remains explicit rather than being guessed.

An original source is not a PLLM method, component, recipe, implementation, or evidence record. Downloading a source does not mean that PLLM built, ran, or reproduced it. SourceRecord validates canonical paper attribution independently from ArtifactLock. Upstream locks begin quarantined, remain external-process/oracle-only when restricted, and cannot become reproduction-verified without pinned artifact and evidence digests. The registry performs no network, import, subprocess, or execution operation. Claims from a source remain attributed to that source and limited to its stated workload and assumptions.

Use the research backlog to distinguish tracked sources, planned work, reference primitives, and structural adaptations. API: pllm.research.