Where, Not What: Compelling Video LLMs to Learn Geometric Causality for 3D-Grounding

• URL: http://arxiv.org/abs/2510.17034v1
• Date: Sun, 19 Oct 2025 22:40:18 GMT
• Title: Where, Not What: Compelling Video LLMs to Learn Geometric Causality for 3D-Grounding
• Authors: Yutong Zhong,
• Abstract summary: We propose a novel training framework called What-Where Representation Re-Forming (W2R2) to tackle this issue.<n>Our approach fundamentally reshapes the model's internal space by designating 2D features as semantic beacons for "What" identification and 3D features as spatial anchors for "Where" localization.<n>Experiments conducted on ScanRefer and ScanQA demonstrate the effectiveness of W2R2, with significant gains in localization accuracy and robustness.
• Score: 0.8883733362171032
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Multimodal 3D grounding has garnered considerable interest in Vision-Language Models (VLMs) \cite{yin2025spatial} for advancing spatial reasoning in complex environments. However, these models suffer from a severe "2D semantic bias" that arises from over-reliance on 2D image features for coarse localization, largely disregarding 3D geometric inputs and resulting in suboptimal fusion performance. In this paper, we propose a novel training framework called What-Where Representation Re-Forming (W2R2) to tackle this issue via disentangled representation learning and targeted shortcut suppression. Our approach fundamentally reshapes the model's internal space by designating 2D features as semantic beacons for "What" identification and 3D features as spatial anchors for "Where" localization, enabling precise 3D grounding without modifying inference architecture. Key components include a dual-objective loss function with an Alignment Loss that supervises fused predictions using adapted cross-entropy for multimodal synergy, and a Pseudo-Label Loss that penalizes overly effective 2D-dominant pseudo-outputs via a margin-based mechanism. Experiments conducted on ScanRefer and ScanQA demonstrate the effectiveness of W2R2, with significant gains in localization accuracy and robustness, particularly in cluttered outdoor scenes.

Related papers

• Debiasing Diffusion Priors via 3D Attention for Consistent Gaussian Splatting [10.320439319212463]
  We propose a novel framework, TD-Attn, which addresses multi-view inconsistency via two key components.<n> TD-Attn has the potential to serve as a universal plugin, significantly enhancing multi-view consistency across 3D tasks.
  arXiv  Detail & Related papers  (2025-12-08T09:33:39Z)
• Beyond Flatlands: Unlocking Spatial Intelligence by Decoupling 3D Reasoning from Numerical Regression [12.590536117486257]
  Existing Vision Language Models (VLMs) struggle to comprehend real-world 3D spatial intelligence.<n> GEODE augments main VLM with two specialized, plug-and-play modules.<n>The synergy of these modules allows our 1.5B parameter model to function as a high-level semantic dispatcher.
  arXiv  Detail & Related papers  (2025-11-14T12:42:07Z)
• Cross-Modal Geometric Hierarchy Fusion: An Implicit-Submap Driven Framework for Resilient 3D Place Recognition [9.411542547451193]
  We propose a novel framework that redefines 3D place recognition through density-agnostic geometric reasoning.<n>Specifically, we introduce an implicit 3D representation based on elastic points, which is immune to the interference of original scene point cloud density.<n>With the aid of these two types of information, we obtain descriptors that fuse geometric information from both bird's-eye view and 3D segment perspectives.
  arXiv  Detail & Related papers  (2025-06-17T07:04:07Z)
• GEAL: Generalizable 3D Affordance Learning with Cross-Modal Consistency [50.11520458252128]
  Existing 3D affordance learning methods struggle with generalization and robustness due to limited annotated data.<n>We propose GEAL, a novel framework designed to enhance the generalization and robustness of 3D affordance learning by leveraging large-scale pre-trained 2D models.<n>GEAL consistently outperforms existing methods across seen and novel object categories, as well as corrupted data.
  arXiv  Detail & Related papers  (2024-12-12T17:59:03Z)
• FILP-3D: Enhancing 3D Few-shot Class-incremental Learning with Pre-trained Vision-Language Models [59.13757801286343]
  Few-shot class-incremental learning aims to mitigate the catastrophic forgetting issue when a model is incrementally trained on limited data.<n>We introduce the FILP-3D framework with two novel components: the Redundant Feature Eliminator (RFE) for feature space misalignment and the Spatial Noise Compensator (SNC) for significant noise.
  arXiv  Detail & Related papers  (2023-12-28T14:52:07Z)
• JOTR: 3D Joint Contrastive Learning with Transformers for Occluded Human Mesh Recovery [84.67823511418334]
  This paper presents 3D JOint contrastive learning with TRansformers framework for handling occluded 3D human mesh recovery. Our method includes an encoder-decoder transformer architecture to fuse 2D and 3D representations for achieving 2D$&$3D aligned results.
  arXiv  Detail & Related papers  (2023-07-31T02:58:58Z)
• Homography Loss for Monocular 3D Object Detection [54.04870007473932]
  A differentiable loss function, termed as Homography Loss, is proposed to achieve the goal, which exploits both 2D and 3D information. Our method yields the best performance compared with the other state-of-the-arts by a large margin on KITTI 3D datasets.
  arXiv  Detail & Related papers  (2022-04-02T03:48:03Z)
• Cylindrical and Asymmetrical 3D Convolution Networks for LiDAR-based Perception [122.53774221136193]
  State-of-the-art methods for driving-scene LiDAR-based perception often project the point clouds to 2D space and then process them via 2D convolution. A natural remedy is to utilize the 3D voxelization and 3D convolution network. We propose a new framework for the outdoor LiDAR segmentation, where cylindrical partition and asymmetrical 3D convolution networks are designed to explore the 3D geometric pattern.
  arXiv  Detail & Related papers  (2021-09-12T06:25:11Z)
• Cylindrical and Asymmetrical 3D Convolution Networks for LiDAR Segmentation [81.02742110604161]
  State-of-the-art methods for large-scale driving-scene LiDAR segmentation often project the point clouds to 2D space and then process them via 2D convolution. We propose a new framework for the outdoor LiDAR segmentation, where cylindrical partition and asymmetrical 3D convolution networks are designed to explore the 3D geometric pat-tern. Our method achieves the 1st place in the leaderboard of Semantic KITTI and outperforms existing methods on nuScenes with a noticeable margin, about 4%.
  arXiv  Detail & Related papers  (2020-11-19T18:53:11Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.