{
  "abstract": "Autoregressive (AR) decoders perform one forward pass per token, so sequential depth grows with output length and the key-value (KV) cache with context. Both costs limit closed-loop control and long structured outputs in which late content constrains early content: the imports and signatures of a program, the pragmas and array declarations of High-Level Synthesis (HLS) C code, the defined terms and cross-references of a contract.\n\nWe propose PR-JEPA (Plan-Realize JEPA), which separates deciding what to generate from emitting it. A Joint-Embedding Predictive Architecture (JEPA) planner, trained with an exponential-moving-average target encoder and VICReg, samples a discrete mode and predicts a continuous plan over the output's structure, such as an abstract syntax tree or a clause outline. A cross-attention bridge conditions a few-step discrete flow-matching realizer that refines all positions in parallel in up to 8 steps, using a closed-form jump kernel tilted by plan-derived constraint energies and clipped to remain valid; a Logit-KL variant integrates a straight path in logit space instead.\n\nWe prove that small VICReg penalties exclude rank-one collapse, that the K-step realizer amplifies plan errors by at most a bounded Lipschitz factor, that plan-mediated hard constraints survive fully parallel updates, that re-anchoring bounds latent error over any horizon, and how to calibrate an uncertainty gate for a target risk level. A roofline model predicts 9 to 21 times lower single-request latency than AR decoding at 8 steps, at about one-eighth of AR's saturated throughput. We pre-register evaluation on general-purpose code, HLS-C, contract drafting, and robot manipulation; end-to-end measurements are pending.",
  "abstract_shown": true,
  "affiliations": [
    "SLING AI Inc."
  ],
  "article_type": null,
  "arxiv": null,
  "author_total": 1,
  "authors": [
    {
      "affs": [
        0
      ],
      "name": "Yeonseok Lee",
      "orcid": "0009-0004-6153-0205"
    }
  ],
  "bibtex": "https://research.draconitic.com/paper/91keswsk4894q.bib",
  "cited_by": 0,
  "doi": "10.5281/zenodo.23498811",
  "fulltext_url": null,
  "id": "91keswsk4894q",
  "issns": [],
  "landing_url": "https://zenodo.org/records/23498811",
  "language": null,
  "licence": "CC-BY-4.0",
  "licence_label": "CC BY 4.0",
  "licence_source": "article",
  "licence_url": "https://creativecommons.org/licenses/by/4.0/",
  "metadata_licence": null,
  "native_subjects": [],
  "openalex": null,
  "pdf_url": "https://zenodo.org/api/records/23498811/files/preprint_261020.pdf/content",
  "pmcid": null,
  "pmid": null,
  "published": "2026-10-20",
  "publisher": null,
  "ris": "https://research.draconitic.com/paper/91keswsk4894q.ris",
  "sources": [
    {
      "first_seen": "2026-11-03T12:20:47Z",
      "native_id": "23498811",
      "source": "zenodo",
      "url": "https://zenodo.org/records/23498811"
    }
  ],
  "subjects": [],
  "title": "PR-JEPA: Decoupling Latent Planning from Few-Step Discrete Realization for Non-Autoregressive Generation of Code, Legal Documents, and Control",
  "title_text": "PR-JEPA: Decoupling Latent Planning from Few-Step Discrete Realization for Non-Autoregressive Generation of Code, Legal Documents, and Control",
  "type": "preprint",
  "updated_at": "2026-11-03T12:20:47Z",
  "url": "https://research.draconitic.com/paper/91keswsk4894q",
  "venue": "Zenodo",
  "versions": []
}