Metraipolis › archive › dc_012bb20fceb4
OPTD: On-Policy Transition Distillation with Consistency-Guided Adaptive Compression for Few-Step Diffusion Language Models — arXiv:2608.02942
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2608.02942
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-31T12:30:50.326Z
- SHA-256 of the archived text
8b1edb150f41dc4abb339d71234cbe4aa277d3b1005104530373564a473c59c6- Stable URL of this page
https://metraipolis.com/documento/dc_012bb20fceb4
How to cite
OPTD: On-Policy Transition Distillation with Consistency-Guided Adaptive Compression for Few-Step Diffusion Language Models — arXiv:2608.02942 — arXiv (en), https://arxiv.org/abs/2608.02942 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_012bb20fceb4 (Metraipolis archive dc_012bb20fceb4, sha256 8b1edb150f41dc4abb339d71234cbe4aa277d3b1005104530373564a473c59c6)
BibTeX
@misc{dc012bb20fceb4,
title = {OPTD: On-Policy Transition Distillation with Consistency-Guided Adaptive Compression for Few-Step Diffusion Language Models — arXiv:2608.02942},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2608.02942},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_012bb20fceb4; sha256 8b1edb150f41dc4abb339d71234cbe4aa277d3b1005104530373564a473c59c6},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.02942. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Diffusion language models (dLLMs) can predict many tokens in parallel, but accurate generation still requires many iterative denoising steps. Few-step distillation accelerates decoding by compressing multiple teacher steps into a single student transition. However, existing methods construct supervision on off-policy trajectories. At inference, the student's early parallel commitments alter the context of later predictions, so the states it actually visits drift away from the supervised ones--precisely when step compression is most aggressive. On-policy distillation is a natural remedy for this mismatch, but it leaves open how far each transition should advance: matching only the teacher's next action limits compression, while indiscriminately merging future actions can violate intermediate dependencies. To address this limitation, we propose OPTD, On-Policy Transition Distillation with consistency-guided adaptive compression. It samples partial states from the few-step student's own trajectories, uses a frozen, question-only teacher to identify outcome-aligned future candidates, and orders them by current-state confidence. The method then selects the longest prefix whose joint commitment preserves the teacher's rollout outcome. A set-bottleneck objective promotes every verified future candidate to the decoder's release threshold, while a frozen-teacher KL anchor regularizes all other active positions. Neither target construction nor training uses a gold response. Across four mathematical reasoning and code-generation benchmarks, OPTD consistently improves the quality--efficiency trade-off and attains the strongest overall quality-constrained AUP among the evaluated few-step baselines. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-05 in cs.CL). Categorie arXiv: Computation and Language (cs.CL). Pagina del lavoro: https://arxiv.org/abs/2608.02942 — 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 2026-10-31T12:17:37.727Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-08-05?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).