Metraipolis › archive › dc_c66ffc85a754

When Parallel Drafter Meets Parallel Speculative Decoding — arXiv:2609.27396

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.27396
License (as archived)
CC0-1.0
Archived
2026-10-08T10:04:58.191Z
SHA-256 of the archived text
d770613735064dd3737f3f6481cba91f5105cc6377968e0c2c32d28deeedfac9
Stable URL of this page
https://metraipolis.com/documento/dc_c66ffc85a754

How to cite

When Parallel Drafter Meets Parallel Speculative Decoding — arXiv:2609.27396 — arXiv (en), https://arxiv.org/abs/2609.27396 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_c66ffc85a754 (Metraipolis archive dc_c66ffc85a754, sha256 d770613735064dd3737f3f6481cba91f5105cc6377968e0c2c32d28deeedfac9)
BibTeX
@misc{dcc66ffc85a754,
  title = {When Parallel Drafter Meets Parallel Speculative Decoding — arXiv:2609.27396},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.27396},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_c66ffc85a754; sha256 d770613735064dd3737f3f6481cba91f5105cc6377968e0c2c32d28deeedfac9},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.27396. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

DSpark-style parallel drafters have made speculative decoding highly effective, yet their draft phase remains serialized on the critical path of every round. Parallel speculative decoding (PSD) overlaps drafting with verification, yet existing methods must guess the accepted prefix and bonus token in advance: a wrong guess reverts the whole batch to serial drafting. We present DPara, a PSD framework that reuses effective parallel drafters yet guarantees backbone--verification overlap in every round, thereby eliminating this probabilistic fallback altogether. While the target verifies, DPara's diffusion backbone precomputes draft representations for every acceptance boundary with the bonus left unspecified; a lightweight autoregressive head then combines the revealed verification outcome with the matching precomputed representation to emit the next round's draft tokens almost instantly---fully parallelizing the dominant backbone forward with verification and leaving only the negligible head cost serial. Experiments on Qwen3-8B and Qwen3-14B across seven math, coding, and chat benchmarks show that DPara achieves average speedups of $3.21\times$ and $3.52\times$ over autoregressive decoding, surpassing the strongest serial and parallel speculative decoding baselines alike.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-24 in cs.CL).
Categorie arXiv: Computation and Language (cs.CL).
Pagina del lavoro: https://arxiv.org/abs/2609.27396 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-08T09:11:49.618Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-09-24?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).