Driving · 2026-10-26
HGWM: Hierarchical Graph-Guided World Model for Zero-Shot Object Navigation via Graph Matching
I-Tak Ieong, Hao Tang
AI-assisted summary based on the paper. Consult the linked source for methods and limitations. Benchmark numbers appear after editor verification.
TL;DR
HGWM is a hierarchical graph-guided world model for zero-shot object navigation. It uses an LLM to generate a category-conditioned room–anchor goal prior and maintains a persistent spatial-semantic graph from RGB-D observations. A coarse stage scores candidate headings with a vision-language model, and a fine stage matches observed objects/relations to the goal prior, fused with phase-dependent arbitration. The authors report improved SR/SPL versus WMNav on HM3D and MP3D.
Why it matters
The authors target zero-shot object navigation where observations may be ambiguous and spatial memory incomplete. HGWM explicitly combines a symbolic relational goal prior with persistent graph memory and uses coarse visual-semantic scoring plus fine, direction-specific graph matching to guide where to look next. The results are reported on HM3D and MP3D, with additional ablations on HM3D v0.2.
Method
- LLM-generated, category-conditioned hierarchical goal graph (rooms, anchors, connections) cached per episode; persistent spatial-semantic memory stores encountered evidence and view associations.
- Two-stage decision: coarse VLM scoring over six headings (0–10) followed by fine symbolic compatibility matching between the current direction’s observed graph evidence and the goal prior, with relation refinement described as heuristic grounding.
- Phase-dependent fusion integrates coarse and fine scores and an LLM direction recommendation; fused values update a stored curiosity map for selecting the next direction and local displacement.
Limitation
The authors report current limitations including simulation-only evidence (no physical-robot results or controlled noise robustness tests), sensitivity evaluation for only two VLM settings on one split, and an assumption of static scenes; they also state performance depends on underlying VLM/LLMs and real-time deployment remains challenging due to VLM inference latency (2.3s per call).
Abstract (from arXiv)
Zero-shot object navigation requires an agent to reason about where a target may be found while retaining evidence from previous observations. We present HGWM, a hierarchical graph-guided world model that combines a category-conditioned room--anchor prior with persistent spatial-semantic memory. A coarse stage uses a vision-language model to score candidate headings, while a fine stage explicitly compares observed objects and relations with the goal prior. Phase-dependent fusion integrates these signals with a language-model recommendation to guide local navigation. HGWM achieves 62.6\% success rate (SR) and 31.5\% success weighted by path length (SPL) on HM3D v0.1, improving over WMNav by 4.5 and 0.3 percentage points, respectively. On MP3D, it achieves 46.8\% SR and 17.9\% SPL, with gains of 1.4 and 0.7 points. On HM3D v0.2, the full model reaches 74.4\% SR and 31.9\% SPL; removing individual components reduces SR by 0.9--2.3 points and SPL by 0.6--1.5 points. These results support the use of relational priors and persistent graph memory for target-directed search.
Related papers
- G$^2$WAM: World Action Models for Efficient and Geometry Grounded Image-Goal Navigation
- World-as-Graph: Relational World Modeling Through Latent Space Graphs
- SkipWorld: Latent Skip Planning with Distilled World Models for Efficient Visual Navigation
- Seeing Beyond the Next Step: World-Model-Guided Human-Like Navigation in Multi-Agent Scenes
- World Models with Configurable Perception
Publication record
Bibliographic evidence describes where and when a record appeared. It does not validate scientific claims.
First public release: 2026-10-23 · Indexed by Pub-AI: 2026-10-26 UTC
HGWM: Hierarchical Graph-Guided World Model for Zero-Shot Object Navigation via Graph Matching
Preprint posted: 2026-10-23 · Original record ↗
- arxiv
- 2610.32554
Explicit source versions (2)
- v1 · 2026-10-23
- v3 · 2026-10-28
- Preprint available
Source date: 2026-10-23 · First detected: 2026-10-27 UTC · Historical baseline
Source evidence (2)
- Official evidence ↗ · atom:published (official metadata)
- Official evidence ↗ · atom:published (official metadata)
- Explicit source version detected · v3
Source date: 2026-10-28 · First detected: 2026-11-08 UTC · Historical baseline
Source evidence (1)
- Official evidence ↗ · atom:id explicit version (official metadata)
Abstract · arxiv
Zero-shot object navigation requires an agent to reason about where a target may be found while retaining evidence from previous observations. We present HGWM, a hierarchical graph-guided world model that combines a category-conditioned room--anchor prior with persistent spatial-semantic memory. A coarse stage uses a vision-language model to score candidate headings, while a fine stage explicitly compares observed objects and relations with the goal prior. Phase-dependent fusion integrates these signals with a language-model recommendation to guide local navigation. HGWM achieves 62.6\% success rate (SR) and 31.5\% success weighted by path length (SPL) on HM3D v0.1, improving over WMNav by 4.5 and 0.3 percentage points, respectively. On MP3D, it achieves 46.8\% SR and 17.9\% SPL, with gains of 1.4 and 0.7 points. On HM3D v0.2, the full model reaches 74.4\% SR and 31.9\% SPL; removing individual components reduces SR by 0.9--2.3 points and SPL by 0.6--1.5 points. These results support the use of relational priors and persistent graph memory for target-directed search.
