{
  "implementation": "The canonical full and assistant-turn exports, exact structured-action scorer, and a pinned Nemotron tokenizer/mask/batch demonstration are available; NLL requires model weights and inference.",
  "reason": "No trained language model weights or inference accelerator were used. Tokenizer-derived lengths and assistant-only masks are reported separately and are not NLL.",
  "status": "not-run"
}