Research map / Vision-language and multimodal / Robot action and navigation
MSGNav: Unleashing the Power of Multi-modal 3D Scene Graph for Zero-Shot Embodied Navigation
Xun Huang, Shijia Zhao, Yunxiang Wang, Xin Lu, Wanfa Zhang, Rongsheng Qu, Weixin Li, Yunhong Wang, Chenglu Wen
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add M3DSG multi-modal scene graph (ablation) | 3D-Mem baseline · GOAT-Bench Val Unseen (first episode per scene) | SR (%): 28.8 -> 43.8 | Improved |
| add VVD viewpoint decision (ablation) | MSGNav (M3DSG) · GOAT-Bench Val Unseen (first episode per scene) | SR (%): 43.8 -> 56.3 | Improved |
| add AVU vocabulary update (ablation) | MSGNav (M3DSG+VVD) · GOAT-Bench Val Unseen (first episode per scene) | SR (%): 56.3 -> 55.3 | Mixed |
| add CLR closed-loop reasoning (ablation) | MSGNav (M3DSG+VVD) · GOAT-Bench Val Unseen (first episode per scene) | SR (%): 56.3 -> 53.2 | Mixed |
| add both AVU and CLR (ablation) | MSGNav (M3DSG+VVD) · GOAT-Bench Val Unseen (first episode per scene) | SR (%): 56.3 -> 60.0 | Improved |
| add relation edges (Concept-graph) to node-only graph (ablation) | MSGNav · GOAT-Bench Val Unseen (first episode per scene) | SR (%): 51.8 -> 56.2 | Improved |
| replace traditional graph with M3DSG (ablation) | MSGNav · GOAT-Bench Val Unseen (first episode per scene) | SR (%): 56.2 -> 60.0 | Improved |
| add VVD at standard threshold 0.25 m (ablation) | MSGNav · GOAT-Bench Val Unseen | SR (%): 33.91 -> 51.97 | Improved |
| add VVD at relaxed threshold 1.00 m (ablation) | MSGNav · GOAT-Bench Val Unseen | SR (%): 62.38 -> 66.52 | Improved |
| use MSGNav instead of MTU3D | MSGNav · GOAT-Bench Val Unseen | SR (%): 47.2 -> 52.0 | Improved |
| use MSGNav instead of WMNav | MSGNav · HM3D-ObjNav | SR (%): 72.2 -> 74.1 | Improved |
| swap GPT-4o for Qwen-VL-Max | MSGNav · GOAT-Bench Val Unseen | SR (%): 51.97 -> 52.79 | No clear effect |
Later papers that compared with this method
1 later accepted paper reports MSGNav as a baseline in their results tables; 0 of them report a better number on most of the results they share with it.
- AMCoNav: Asynchronous Multi-module Collaborative Framework for Embodied Visual Navigation (ECCV 2026 · better on 1 of 2 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- BeliefMapNav: 3D Voxel-Based Belief Map for Zero-Shot Object Navigation (NeurIPS 2025 · similarity 0.90)
- SG-Nav: Online 3D Scene Graph Prompting for LLM-based Zero-shot Object Navigation (NeurIPS 2024 · similarity 0.89)
- UniGoal: Towards Universal Zero-shot Goal-oriented Navigation (CVPR 2025 · similarity 0.89)
- Hierarchical Semantic-Augmented Navigation: Optimal Transport and Graph-Driven Reasoning for Vision-Language Navigation (NeurIPS 2025 · similarity 0.89)
- D3D-VLP: Dynamic 3D Vision-Language-Planning Model for Embodied Grounding and Navigation (CVPR 2026 · similarity 0.88)
- VoroNav: Voronoi-based Zero-shot Object Navigation with Large Language Model (ICML 2024 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
