---
title: "CoverPrune：基于最优传输的覆盖率驱动3D视觉语言模型令牌剪枝方法"
source: "arXiv AI/CL/LG（论文）"
source_tier: "t2"
url: "https://arxiv.org/abs/2608.13226v1"
permalink: "https://feed.astralor.com/items/itm_d6d7376d7de437b40475b9e9"
published: "2026-08-13T13:29:15.000Z"
score: 72
channel: "tech"
tags: ["3D视觉语言模型","arXiv","CoverPrune","令牌剪枝","最优传输","空间推理"]
---

# CoverPrune：基于最优传输的覆盖率驱动3D视觉语言模型令牌剪枝方法

## AI 点评

> 今日最值得看的点在于：它把剪枝逻辑从“选不同”颠覆为“保证据”，这正是3D任务比2D更需要的东西。对研究3D VLM推理加速或者做多模态模型轻量化的人，这个覆盖率视角可能会打开一条新路。

## 速览

3D视觉语言模型（3D VLM）在空间推理上表现优异，但海量视觉令牌导致推理计算瓶颈。现有剪枝方法多基于多样性选择，在3D环境中易丢弃关键原型令牌，破坏多视角一致性和几何结构。为此，研究团队提出CoverPrune，一种无需训练的框架，将推理时令牌剪枝建模为最优传输问题，并设计特征-空间-时间传输代价及空间引导贪心选择算法近似求解，同时推出加速版CoverPrune-Lite。实验表明，该方法在多个3D空间推理基准上取得领先的令牌效率，即使在高压缩比下仍保持稳健性能。

## 关键点

- 提出将3D VLM令牌剪枝从“最大化多样性”转向“保留视觉证据覆盖”的新范式。
- 设计基于最优传输的无训练框架CoverPrune，并推出低开销加速版CoverPrune-Lite。
- 在多个3D空间推理基准上实现领先的令牌效率，高剪枝率下依然保持稳健推理性能。

## 全文

While 3D Vision-Language Models (3D VLMs) have demonstrated remarkable spatial reasoning capabilities, they suffer from massive visual token counts that create severe computational bottlenecks during inference. Existing token pruning methods primarily rely on diversity-based selection, discarding similar tokens to maximize dispersion. However, in 3D environments, this approach frequently drops representative prototype tokens in favor of outliers, breaking the multi-view consistencies and geometric structures essential for spatial reasoning. In this paper, we propose a paradigm shift for 3D VLM token pruning: from maximizing diversity to preserving visual evidence coverage. We introduce CoverPrune, a training-free framework that formulates inference-time token pruning as an Optimal Transport (OT) problem. To overcome the intractable combinatorial subset selection inherent in this formulation, we design the Feature-Spatial-Temporal (FST) transport cost and target capacity, along with an efficient Spatial-Guided Greedy Selection (SGS) algorithm to approximate the OT objective. Furthermore, we propose CoverPrune-Lite, an accelerated variant utilizing spatially structured local matching for minimal overhead. Extensive experiments across multiple 3D visual-spatial reasoning benchmarks demonstrate that our methods achieve state-of-the-art token efficiency, maintaining robust reasoning performance even under highly aggressive pruning budgets. Visit our project website at https://github.com/Brucess/CoverPrune.

---

引用前请核对原文;本文由星际流动整理:https://feed.astralor.com/items/itm_d6d7376d7de437b40475b9e9