Metraipolis › archive › dc_08b6747893bc
Perception Before Reasoning: Dynamic Latent Reasoning for Video Understanding and Question Answering — arXiv:2608.04124
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2608.04124
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-31T11:13:34.064Z
- SHA-256 of the archived text
11e6ee30d6b6617e213bc63e339cb32cbd44bc7ffcdcb657592cc6be7faf3127- Stable URL of this page
https://metraipolis.com/documento/dc_08b6747893bc
How to cite
Perception Before Reasoning: Dynamic Latent Reasoning for Video Understanding and Question Answering — arXiv:2608.04124 — arXiv (en), https://arxiv.org/abs/2608.04124 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_08b6747893bc (Metraipolis archive dc_08b6747893bc, sha256 11e6ee30d6b6617e213bc63e339cb32cbd44bc7ffcdcb657592cc6be7faf3127)
BibTeX
@misc{dc08b6747893bc,
title = {Perception Before Reasoning: Dynamic Latent Reasoning for Video Understanding and Question Answering — arXiv:2608.04124},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2608.04124},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_08b6747893bc; sha256 11e6ee30d6b6617e213bc63e339cb32cbd44bc7ffcdcb657592cc6be7faf3127},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.04124. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Video question answering requires models to ground language queries in visual evidence and, when necessary, reason over that evidence across time. Existing methods typically rely on long textual chain-of-thought rationales, even though many questions can be answered as soon as the relevant object, action, or frame is localized. We propose Dynamic Latent Reasoning (DyLaR), which first grounds a question in a short block of perception latents (continuous hidden states that encode query-relevant visual evidence), and then adaptively decides whether to append reasoning latents (continuous thoughts that reason over this evidence in latent space) before answering. DyLaR learns this behavior by grounding perception latents in verified visual evidence and distilling verified rationales into reasoning latents, followed by reinforcement learning that further refines when to reason. Across nine video benchmarks and four multimodal language model backbones, DyLaR improves average accuracy over same-backbone baselines while generating fewer than 20 tokens per query. On Qwen3-VL-4B, for example, DyLaR improves average accuracy over Qwen3-VL-4B-Thinking from 54.0 to 58.2 while reducing response length from 1,220.7 to 18.5 tokens per query. Ablations further show that grounded perception latents, rationale-supervised reasoning latents, and adaptive routing each improve accuracy. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-06 in cs.AI). Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV) ; Artificial Intelligence (cs.AI). Pagina del lavoro: https://arxiv.org/abs/2608.04124 — 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 2026-10-31T11:02:13.482Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-08-06?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).