Metraipolis › archive › dc_fb23a2022eb9
Routing by Reasoning Need: Trajectory-Aware Decoding Control for Diffusion Vision-Language Models — arXiv:2609.11315
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.11315
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-16T05:24:03.309Z
- SHA-256 of the archived text
1e7e3742eee31e02ebc3b0361b79ade929ad711e7c2b5c2e28bac04a6a0293e9- Stable URL of this page
https://metraipolis.com/documento/dc_fb23a2022eb9
How to cite
Routing by Reasoning Need: Trajectory-Aware Decoding Control for Diffusion Vision-Language Models — arXiv:2609.11315 — arXiv (en), https://arxiv.org/abs/2609.11315 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_fb23a2022eb9 (Metraipolis archive dc_fb23a2022eb9, sha256 1e7e3742eee31e02ebc3b0361b79ade929ad711e7c2b5c2e28bac04a6a0293e9)
BibTeX
@misc{dcfb23a2022eb9,
title = {Routing by Reasoning Need: Trajectory-Aware Decoding Control for Diffusion Vision-Language Models — arXiv:2609.11315},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.11315},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_fb23a2022eb9; sha256 1e7e3742eee31e02ebc3b0361b79ade929ad711e7c2b5c2e28bac04a6a0293e9},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.11315. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Diffusion vision-language models generate answers through iterative refinement, exposing intermediate answer trajectories that can be inspected and controlled at inference time. However, this controllability creates a reasoning-need mismatch, where a universal generation length is applied to questions with different reasoning demands. Visually closed questions may be harmed by continued refinement after a stable answer has formed, whereas reasoning-sensitive questions may be harmed by premature commitment. We formulate this problem as reasoning-budget mismatch and study it in LLaDA-V. Rather than choosing a universal generation length, our training-free controller routes each example to early commitment, baseline preservation, or reasoning-supportive decoding using trajectory signals from answer closure, commitment evidence, and representation revision pressure, without using ground-truth answers. Across answer-focused, mixed-reasoning, and CoT-sensitive benchmarks, routed control improves robustness over fixed long decoding, pure short decoding, and single-rule interventions. The gains are not explained by shorter outputs alone. Answer-closed examples often benefit from commitment, whereas CoT-sensitive examples require preserving or supporting intermediate reasoning. Taken together, these results suggest diffusion VLM decoding should route inference-time control by the state suggested by the observed trajectory instead of relying on a universal decoding length. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-11 in cs.AI). Categorie arXiv: Artificial Intelligence (cs.AI). Pagina del lavoro: https://arxiv.org/abs/2609.11315 — 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 2026-10-16T05:17:52.004Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-11?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).