跳到正文
研踪PaperTrail
← 研踪 PaperTrail

arXiv:2611.22577 · v1

Semantic-Guided Geometric Decoding for Token-Efficient Visual Grounding

首次提交 2026-11-17 19:02
版本更新 2026-11-17 19:02
雷达观测 2026-11-18 11:33
arXiv ↗PDF ↗

摘要

Multimodal large language models (MLLMs) provide a unified interface for language-conditioned visual perception. In pixel-to-sequence grounding, autoregressive coordinate generation requires multiple sequential decoding steps per box, increasing inference cost with the number of instances. Pixel-to-embedding methods reduce this overhead but lack explicit spatial encoding when using a shared special token. We introduce SemGeo, a token-efficient visual grounding framework that represents each instance with a single grid token for coarse localization. A geometric decoder reuses visual features and hidden states to recover precise discrete box coordinates in parallel. To train proposal generation and box localization together, we introduce Gaussian soft labels that encode coordinate distances during supervised fine-tuning (SFT). Reinforcement post-training further combines box-set rewards with coordinate supervision on model-generated proposals. This stage trains the language model and geometric decoder on their own predictions to mitigate the mismatch between teacher-forced training and inference. Across eight benchmarks spanning object detection, referring object detection, and document layout grounding, SemGeo achieves the highest F1@mIou among compared MLLMs on six benchmarks. It also delivers 2.07×2.07\times and 1.25×1.25\times the mean box-generation throughput of representative grounding baselines under vLLM and Transformers, respectively, while achieving higher accuracy in both comparisons.

作者备注:10 pages, 4 figures, 5 tables

AI 论文解读

基于摘要 · 非全文审阅

此版本尚未生成 AI 解读。解读由后台任务生成并按论文版本保存,登录后共享已有结果。

版本差异

目前只收录了一个版本。收录下一版后,可以对照标题、作者、分类和摘要。

观测到的版本

相邻记录可能跳过中间版本;首次发现 v2 及以上也会标为版本更新。

v1 · 2026-11-17 19:02 · 最新观测版本

Semantic-Guided Geometric Decoding for Token-Efficient Visual Grounding

Multimodal large language models (MLLMs) provide a unified interface for language-conditioned visual perception. In pixel-to-sequence grounding, autoregressive coordinate generation requires multiple sequential decoding steps per box, increasing inference cost with the number of instances. Pixel-to-embedding methods reduce this overhead but lack explicit spatial encoding when using a shared special token. We introduce SemGeo, a token-efficient visual grounding framework that represents each instance with a single grid token for coarse localization. A geometric decoder reuses visual features and hidden states to recover precise discrete box coordinates in parallel. To train proposal generation and box localization together, we introduce Gaussian soft labels that encode coordinate distances during supervised fine-tuning (SFT). Reinforcement post-training further combines box-set rewards with coordinate supervision on model-generated proposals. This stage trains the language model and geometric decoder on their own predictions to mitigate the mismatch between teacher-forced training and inference. Across eight benchmarks spanning object detection, referring object detection, and document layout grounding, SemGeo achieves the highest F1@mIou among compared MLLMs on six benchmarks. It also delivers 2.07×2.07\times and 1.25×1.25\times the mean box-generation throughput of representative grounding baselines under vLLM and Transformers, respectively, while achieving higher accuracy in both comparisons.

查看此版本解读 →查看此版本原文 ↗