Metraipolis › archive › dc_aef880e2dd4e

Skip the Talk, Re-Focus on Vision: Latent Reasoning for Reasoning Segmentation in Multimodal Large Language Models — arXiv:2609.30783

Source: arXiv (en)
Original: https://arxiv.org/abs/2609.30783
License (as archived): CC0-1.0
Archived: 2026-10-08T01:39:30.028Z
SHA-256 of the archived text: 1a4bd367b0c60817612d629d90776ee94c6a95f357cb9b39dd3aa6db29e381d7
Stable URL of this page: https://metraipolis.com/documento/dc_aef880e2dd4e

How to cite

Skip the Talk, Re-Focus on Vision: Latent Reasoning for Reasoning Segmentation in Multimodal Large Language Models — arXiv:2609.30783 — arXiv (en), https://arxiv.org/abs/2609.30783 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_aef880e2dd4e (Metraipolis archive dc_aef880e2dd4e, sha256 1a4bd367b0c60817612d629d90776ee94c6a95f357cb9b39dd3aa6db29e381d7)

BibTeX

@misc{dcaef880e2dd4e,
  title = {Skip the Talk, Re-Focus on Vision: Latent Reasoning for Reasoning Segmentation in Multimodal Large Language Models — arXiv:2609.30783},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.30783},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_aef880e2dd4e; sha256 1a4bd367b0c60817612d629d90776ee94c6a95f357cb9b39dd3aa6db29e381d7},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.30783. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Reasoning segmentation aims to interpret implicit textual queries and enable fine-grained visual perception, which is critical for applications such as human-computer interaction and embodied agents. Existing methods typically generate explicit Chain-of-Thought (CoT) by multimodal large language models (MLLMs) before localizing the target. Although intuitive, such explicit verbal reasoning introduces substantial attention interference: redundant textual tokens disrupt attention during perception-token generation and also increase the effective distance between visual tokens. To address this issue, we propose LIRSeg, which fully replaces explicit CoT with a compact set of learnable latent tokens for reasoning segmentation. LIRSeg is trained in two stages: spatial alignment grounds the latent tokens in object-relevant visual evidence, and GRPO further optimizes them with segmentation rewards. To make these compact latent tokens more informative, we introduce three complementary mechanisms from an information perspective: extreme-advantage sampling for selecting informative training signals, decoupled exploration-stability updates for learning complementary representations, and latent diversity amplification for preventing representational collapse. Extensive experiments on benchmarks demonstrate that LIRSeg consistently improves both segmentation accuracy and reasoning efficiency. Compared with the VisionReasoner baseline, LIRSeg achieves absolute gIoU improvements of 4.9% on ReasonSeg, 7.1% on MUSE, and 4.7% on MMR, while achieving a approximately 16x reduction in reasoning tokens. Code is available in supplementary materials.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-28 in cs.AI).
Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV) ; Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2609.30783 — 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 2026-10-08T01:12:14.480Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-28?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).