Metraipolis › archive › dc_4d82f2864263
Draft in Parallel, Condition Through Depth: Adjacent Causal Injection for Speculative Decoding — arXiv:2609.36173
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.36173
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-07T16:25:25.729Z
- SHA-256 of the archived text
149f40c3c2d6afe48cee2b646d416371c4c3bd231ae11e7d0a7e5ca2bd943f59- Stable URL of this page
https://metraipolis.com/documento/dc_4d82f2864263
How to cite
Draft in Parallel, Condition Through Depth: Adjacent Causal Injection for Speculative Decoding — arXiv:2609.36173 — arXiv (en), https://arxiv.org/abs/2609.36173 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_4d82f2864263 (Metraipolis archive dc_4d82f2864263, sha256 149f40c3c2d6afe48cee2b646d416371c4c3bd231ae11e7d0a7e5ca2bd943f59)
BibTeX
@misc{dc4d82f2864263,
title = {Draft in Parallel, Condition Through Depth: Adjacent Causal Injection for Speculative Decoding — arXiv:2609.36173},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.36173},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_4d82f2864263; sha256 149f40c3c2d6afe48cee2b646d416371c4c3bd231ae11e7d0a7e5ca2bd943f59},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.36173. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Parallel speculative drafting generates multiple candidates in one backbone pass, but independent token selection can produce inconsistent continuations that shorten the accepted prefix. Existing methods mostly leave conditional decoding to a lightweight module after the backbone, which limits the flow of predecessor information to successors. Our analysis of DFlash shows that early positions already form recoverable predictions in shallow layers, and that accurate adjacent predecessors help successors more when they enter earlier. We therefore propose DSpine, a drafter with causal conditioning injection throughout the backbone: at every layer, gated adjacent injection writes each predecessor's predicted feature into its successor, so the causal conditioning chain unfolds over network depth while all positions update in parallel. A unified transfer space built from the target model's output embeddings unifies layer-wise injection with predecessor-conditioned decoding, and layer-wise output-embedding supervision promotes the formation of predicted features in shallow layers. Fused kernels and a transition cache execute both efficiently in parallel within SGLang. Across seven math, code, and chat benchmarks, DSpine achieves the longest acceptance length at both temperatures on Qwen3-4B and Qwen3-8B. At temperature zero on Qwen3-8B, it raises the seven-benchmark mean from DFlash's 3.77 to 4.82 (+27.8%); in SGLang serving tests, it delivers 23.3% higher throughput than DFlash on average. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-30 in cs.AI). Categorie arXiv: Machine Learning (cs.LG) ; Artificial Intelligence (cs.AI). Pagina del lavoro: https://arxiv.org/abs/2609.36173 — 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 2026-10-07T15:54:53.785Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-30?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).