Reproductions

PLLM reproductions begin with public papers and locked source records. Contributors then implement the method independently in PLLM. Upstream repositories can help with attribution and comparison, but PLLM does not build them into the runtime or use them as dependencies.

The record reports each step separately: source acquired, reference behavior implemented, primitive fidelity tested, native path implemented, model coverage completed, runtime execution available, benchmark matched, privacy assessed, and limitations reviewed. One summary label must not hide a missing step.

Follow the clean-room workflow and agent checklist.