Metraipolis › archive › dc_5458a892385f
GleanVID: Complementary Token Selection for Efficient Video Large Language Models — arXiv:2609.37042
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.37042
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-07T17:32:45.241Z
- SHA-256 of the archived text
73e205bff14830ecb655d781a16031f879df7a0d58a98f4540e6e9938e60be0b- Stable URL of this page
https://metraipolis.com/documento/dc_5458a892385f
How to cite
GleanVID: Complementary Token Selection for Efficient Video Large Language Models — arXiv:2609.37042 — arXiv (en), https://arxiv.org/abs/2609.37042 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_5458a892385f (Metraipolis archive dc_5458a892385f, sha256 73e205bff14830ecb655d781a16031f879df7a0d58a98f4540e6e9938e60be0b)
BibTeX
@misc{dc5458a892385f,
title = {GleanVID: Complementary Token Selection for Efficient Video Large Language Models — arXiv:2609.37042},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.37042},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_5458a892385f; sha256 73e205bff14830ecb655d781a16031f879df7a0d58a98f4540e6e9938e60be0b},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.37042. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Video Large Language Models (VideoLLMs) have achieved strong video understanding capabilities but incur substantial inference overhead due to the large number of visual tokens. Existing VideoLLM token compression methods largely rely on selection-independent scoring, overlooking cross-frame complementarity and consequently retaining redundant evidence across frames. Instead, we view video token selection as a progressive evidence accumulation process. It aims to retain visual evidence that is individually informative and collectively complementary under a limited token budget. Building on this insight, we introduce GleanVID, a training-free inference acceleration framework for VideoLLMs. Specifically, GleanVID first allocates the global token budget across frames according to temporal novelty and then selects tokens by jointly considering local representativeness and subspace complementarity, thereby preserving richer and less redundant visual evidence. Extensive experiments across diverse VideoLLMs and benchmarks demonstrate that GleanVID consistently achieves state-of-the-art performance. Notably, with only 25% of visual tokens, GleanVID preserves 98.6% of Qwen3-VL's original performance while reducing its prefill latency by 44.7%. On LLaVA-OV-7B, GleanVID at a 25% retention ratio even slightly surpasses the original model. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-30 in cs.LG). Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV) ; Machine Learning (cs.LG). Pagina del lavoro: https://arxiv.org/abs/2609.37042 — 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 2026-10-07T16:59:48.518Z dalla pagina https://arxiv.org/catchup/cs.LG/2026-09-30?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).