Error-Driven Scene Editing for 3D Grounding in Large Language Models

• URL: http://arxiv.org/abs/2511.14086v1
• Date: Tue, 18 Nov 2025 03:13:29 GMT
• Title: Error-Driven Scene Editing for 3D Grounding in Large Language Models
• Authors: Yue Zhang, Zun Wang, Han Lin, Jialu Li, Jianing Yang, Yonatan Bitton, Idan Szpektor, Mohit Bansal,
• Abstract summary: Despite recent progress in 3D-LLMs, they remain limited in accurately grounding language to visual and spatial elements in 3D environments.<n>This limitation stems in part from training data that focuses on language reasoning rather than spatial understanding due to scarce 3D resources.<n>We propose 3D scene editing as a key mechanism to generate precise visual counterfactuals that mitigate these biases.
• Score: 71.41120775319088
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Despite recent progress in 3D-LLMs, they remain limited in accurately grounding language to visual and spatial elements in 3D environments. This limitation stems in part from training data that focuses on language reasoning rather than spatial understanding due to scarce 3D resources, leaving inherent grounding biases unresolved. To address this, we propose 3D scene editing as a key mechanism to generate precise visual counterfactuals that mitigate these biases through fine-grained spatial manipulation, without requiring costly scene reconstruction or large-scale 3D data collection. Furthermore, to make these edits targeted and directly address the specific weaknesses of the model, we introduce DEER-3D, an error-driven framework following a structured "Decompose, Diagnostic Evaluation, Edit, and Re-train" workflow, rather than broadly or randomly augmenting data as in conventional approaches. Specifically, upon identifying a grounding failure of the 3D-LLM, our framework first diagnoses the exact predicate-level error (e.g., attribute or spatial relation). It then executes minimal, predicate-aligned 3D scene edits, such as recoloring or repositioning, to produce targeted counterfactual supervision for iterative model fine-tuning, significantly enhancing grounding accuracy. We evaluate our editing pipeline across multiple benchmarks for 3D grounding and scene understanding tasks, consistently demonstrating improvements across all evaluated datasets through iterative refinement. DEER-3D underscores the effectiveness of targeted, error-driven scene editing in bridging linguistic reasoning capabilities with spatial grounding in 3D LLMs.

Related papers

• PatchAlign3D: Local Feature Alignment for Dense 3D Shape understanding [67.15800065888887]
  Current foundation models for 3D shapes excel at global tasks (retrieval, classification) but transfer poorly to local part-level reasoning.<n>We introduce an encoder-only 3D model that produces language-aligned patch-level features directly from point clouds.<n>Our 3D encoder achieves zero-shot 3D part segmentation with fast single-pass inference without any test-time multi-view rendering.
  arXiv  Detail & Related papers  (2026-01-05T18:55:45Z)
• ZING-3D: Zero-shot Incremental 3D Scene Graphs via Vision-Language Models [0.0]
  ZING-3D is a framework that generates a rich semantic representation of a 3D scene in a zero-shot manner.<n>It also enables incremental updates and geometric grounding in 3D space, making it suitable for downstream robotics applications.<n>Our experiments on scenes from the Replica and HM3D dataset show that ZING-3D is effective at capturing spatial and relational knowledge without the need of task-specific training.
  arXiv  Detail & Related papers  (2025-10-24T00:52:33Z)
• 3DSceneEditor: Controllable 3D Scene Editing with Gaussian Splatting [31.98493679748211]
  We propose 3DSceneEditor, a fully 3D-based paradigm for real-time, precise editing of 3D scenes using Gaussian Splatting.<n>Unlike conventional methods, 3DSceneEditor operates through a streamlined 3D pipeline, enabling direct manipulation of Gaussians for efficient, high-quality edits.
  arXiv  Detail & Related papers  (2024-12-02T15:03:55Z)
• EditRoom: LLM-parameterized Graph Diffusion for Composable 3D Room Layout Editing [114.14164860467227]
  We propose EditRoom, a framework capable of executing a variety of layout edits through natural language commands.<n>Specifically, EditRoom leverages Large Language Models (LLMs) for command planning and generates target scenes.<n>We have developed an automatic pipeline to augment existing 3D scene datasets and introduced EditRoom-DB, a large-scale dataset with 83k editing pairs.
  arXiv  Detail & Related papers  (2024-10-03T17:42:24Z)
• Rethinking Open-Vocabulary Segmentation of Radiance Fields in 3D Space [10.49905491984899]
  We redefine the problem to segment the 3D volume and propose the following methods for better 3D understanding.<n>We directly supervise the 3D points to train the language embedding field, unlike previous methods that anchor supervision at 2D pixels.<n>We transfer the learned language field to 3DGS, achieving the first real-time rendering speed without sacrificing training time or accuracy.
  arXiv  Detail & Related papers  (2024-08-14T09:50:02Z)
• Reasoning3D -- Grounding and Reasoning in 3D: Fine-Grained Zero-Shot Open-Vocabulary 3D Reasoning Part Segmentation via Large Vision-Language Models [20.277479473218513]
  We introduce a new task: Zero-Shot 3D Reasoning for parts searching and localization for objects. We design a simple baseline method, Reasoning3D, with the capability to understand and execute complex commands. We show that Reasoning3D can effectively localize and highlight parts of 3D objects based on implicit textual queries.
  arXiv  Detail & Related papers  (2024-05-29T17:56:07Z)
• Grounded 3D-LLM with Referent Tokens [58.890058568493096]
  We propose Grounded 3D-LLM to consolidate various 3D vision tasks within a unified generative framework. The model uses scene referent tokens as special noun phrases to reference 3D scenes. Per-task instruction-following templates are employed to ensure natural and diversity in translating 3D vision tasks into language formats.
  arXiv  Detail & Related papers  (2024-05-16T18:03:41Z)
• CoT3DRef: Chain-of-Thoughts Data-Efficient 3D Visual Grounding [23.885017062031217]
  3D visual grounding is the ability to localize objects in 3D scenes conditioned by utterances. Most existing methods devote the referring head to localize the referred object directly, causing failure in complex scenarios. We formulate the 3D visual grounding problem as a sequence-to-sequence Seq2Seq task by first predicting a chain of anchors and then the final target.
  arXiv  Detail & Related papers  (2023-10-10T00:07:25Z)
• Four Ways to Improve Verbo-visual Fusion for Dense 3D Visual Grounding [56.00186960144545]
  3D visual grounding is the task of localizing the object in a 3D scene which is referred by a description in natural language. We propose a dense 3D grounding network, featuring four novel stand-alone modules that aim to improve grounding performance.
  arXiv  Detail & Related papers  (2023-09-08T19:27:01Z)
• Visual Localization using Imperfect 3D Models from the Internet [54.731309449883284]
  This paper studies how imperfections in 3D models affect localization accuracy. We show that 3D models from the Internet show promise as an easy-to-obtain scene representation.
  arXiv  Detail & Related papers  (2023-04-12T16:15:05Z)
• Delving into Localization Errors for Monocular 3D Object Detection [85.77319416168362]
  Estimating 3D bounding boxes from monocular images is an essential component in autonomous driving. In this work, we quantify the impact introduced by each sub-task and find the localization error' is the vital factor in restricting monocular 3D detection.
  arXiv  Detail & Related papers  (2021-03-30T10:38:01Z)

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.