Metraipolis › archive › dc_aa5eb8bd5040
BV Loss: Block Verification-Aware Loss for Block Diffusion Speculative Decoding — arXiv:2609.34832
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.34832
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-07T20:58:13.548Z
- SHA-256 of the archived text
bb818b78dd1d500f25ef893dd96b9551246dbb5b197143b962c08cf089324f69- Stable URL of this page
https://metraipolis.com/documento/dc_aa5eb8bd5040
How to cite
BV Loss: Block Verification-Aware Loss for Block Diffusion Speculative Decoding — arXiv:2609.34832 — arXiv (en), https://arxiv.org/abs/2609.34832 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_aa5eb8bd5040 (Metraipolis archive dc_aa5eb8bd5040, sha256 bb818b78dd1d500f25ef893dd96b9551246dbb5b197143b962c08cf089324f69)
BibTeX
@misc{dcaa5eb8bd5040,
title = {BV Loss: Block Verification-Aware Loss for Block Diffusion Speculative Decoding — arXiv:2609.34832},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.34832},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_aa5eb8bd5040; sha256 bb818b78dd1d500f25ef893dd96b9551246dbb5b197143b962c08cf089324f69},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.34832. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Diffusion drafters accelerate speculative decoding by proposing multiple tokens in parallel. Despite recent advances in speculative decoding through sequence-level drafting and verification, existing training objectives remain largely designed around token-level verification. To address this mismatch, we introduce Block Verification-aware loss (BV loss), a training objective designed to maximize the expected acceptance length of a drafted sequence. BV loss is directly derived from the block verification acceptance rule, providing a principled connection between the drafter training objective and the inference-time verification mechanism at the sequence level. Across math, code, and chat benchmarks, BV loss increases the mean number of tokens accepted per verification call under block verification by 13.0--21.0\% over cross-entropy loss training for DFlash and DSpark with Qwen3-4B and Qwen3-8B without changing the inference procedure. BV loss also outperforms tokenwise acceptance objectives such as TV loss and LK loss, and its gains extend to token verification and greedy decoding. These results demonstrate the benefit of training block diffusion drafters with an objective aligned with sequence-level verification, rather than optimizing each token independently. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-29 in cs.AI). Categorie arXiv: Artificial Intelligence (cs.AI) ; Computation and Language (cs.CL). Pagina del lavoro: https://arxiv.org/abs/2609.34832 — 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 2026-10-07T20:29:01.681Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-29?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).