Metraipolis › archive › dc_e1a759dd37ca
Reliable Parallel Decoding in Masked Diffusion Language Models — arXiv:2609.36452
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.36452
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-07T16:25:50.047Z
- SHA-256 of the archived text
1f5b6cd4249896073455a74a0b6ba052fc54f1e7c57015ae2ae9a1a911204458- Stable URL of this page
https://metraipolis.com/documento/dc_e1a759dd37ca
How to cite
Reliable Parallel Decoding in Masked Diffusion Language Models — arXiv:2609.36452 — arXiv (en), https://arxiv.org/abs/2609.36452 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_e1a759dd37ca (Metraipolis archive dc_e1a759dd37ca, sha256 1f5b6cd4249896073455a74a0b6ba052fc54f1e7c57015ae2ae9a1a911204458)
BibTeX
@misc{dce1a759dd37ca,
title = {Reliable Parallel Decoding in Masked Diffusion Language Models — arXiv:2609.36452},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.36452},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_e1a759dd37ca; sha256 1f5b6cd4249896073455a74a0b6ba052fc54f1e7c57015ae2ae9a1a911204458},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.36452. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Masked diffusion language models (MDLMs) can generate text efficiently by predicting multiple masked tokens in parallel, but predictions from the same forward pass are not necessarily reliable when committed together. We study when parallel commitment is reliable. Our diagnostics show that confidence alone does not determine a reliable commitment order: confident predictions near the end of the sequence can fix an answer before its supporting computations are established, and downstream predictions become less reliable as the uncertainty of their upstream context grows. At the same time, a single forward pass can already resolve several masked tokens, and predictions that remain stable across the final layers are more likely to be correct. Based on these findings, we propose Reliable Parallel Decoding (RPD), a training-free method that selects candidates by layerwise prediction stability and final confidence, and commits them under a cumulative entropy budget over their preceding masked positions. RPD defers predictions with uncertain upstream context while committing the remaining candidates in parallel, without relying on a fixed block schedule. Across mathematical reasoning and code generation benchmarks on LLaDA and Dream, RPD achieves the highest decoding throughput among the evaluated methods while maintaining or improving accuracy. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-30 in cs.AI). Categorie arXiv: Computation and Language (cs.CL) ; Artificial Intelligence (cs.AI). Pagina del lavoro: https://arxiv.org/abs/2609.36452 — 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 2026-10-07T15:54:53.838Z 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).