Metraipolis › archive › dc_a7b225cfcd5b

Enhancing Long-Video VLM Embeddings with Query-Aware Streaming Latent Reasoning — arXiv:2610.04864

Source
arXiv
Original
https://arxiv.org/abs/2610.04864
License (as archived)
CC0-1.0
Archived
2026-10-06T05:26:28.210Z
SHA-256 of the archived text
3fc485615f4fe83e25e7e1440bc6ebf3b9b1012295372f4d75bbbaba5972c37b
Stable URL of this page
https://metraipolis.com/documento/dc_a7b225cfcd5b

How to cite

Enhancing Long-Video VLM Embeddings with Query-Aware Streaming Latent Reasoning — arXiv:2610.04864 — arXiv, https://arxiv.org/abs/2610.04864 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_a7b225cfcd5b (Metraipolis archive dc_a7b225cfcd5b, sha256 3fc485615f4fe83e25e7e1440bc6ebf3b9b1012295372f4d75bbbaba5972c37b)
BibTeX
@misc{dca7b225cfcd5b,
  title = {Enhancing Long-Video VLM Embeddings with Query-Aware Streaming Latent Reasoning — arXiv:2610.04864},
  howpublished = {arXiv},
  url = {https://arxiv.org/abs/2610.04864},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_a7b225cfcd5b; sha256 3fc485615f4fe83e25e7e1440bc6ebf3b9b1012295372f4d75bbbaba5972c37b},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv, https://arxiv.org/abs/2610.04864. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Long-video embedding requires capturing sparse query-relevant evidence under a limited visual-token budget. Uniform sampling can miss brief events in videos spanning minutes or hours, whereas encoding more frames in a single context increases memory and computation. We introduce \textbf{Query-Aware Streaming Latent Reasoning} (QASLR), a post-training framework that accumulates evidence across clips while keeping the embedding size fixed. QASLR selects a bounded set of frames, restores their temporal order, and processes them clip by clip with a vision-language backbone. A compact set of persistent think tokens cross-attends to each clip's features, while an embed token reads out a normalized representation after every update. This design integrates evidence across multiple backbone calls without requiring all selected frames to share a single context. Training combines final contrastive learning, step-wise contrastive supervision, and final-embedding self-distillation. Intermediate supervision trains partial-video readouts for retrieval, while self-distillation regularizes them toward the final representation. Query-aware selection produces query-conditioned representations for candidate-set scoring and reranking, whereas query-independent selection enables reusable corpus indexing. Under the full training recipe, HourVideo retrieval Hit@1 increases from 54.2 to 70.7 and from 57.4 to 72.8 for 2B and 8B Qwen3-VL-Embedding backbones, respectively. Gains extend to the evaluated moment-retrieval and video-QA tasks, and the streaming head transfers to a second Qwen-family embedding backbone. These results support streaming latent aggregation as an effective approach to integrating long-video evidence into fixed-dimensional representations.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno).
Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV) ; Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2610.04864 — 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 2026-10-06T05:24:05.921Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).