Research map / Vision-language and multimodal / Robot action and navigation
ZSON: Zero-Shot Object-Goal Navigation using Multimodal Goal Embeddings
Arjun Majumdar, Gunjan Aggarwal, Bhavika Suresh Devnani, Judy Hoffman, Dhruv Batra
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 |
|---|---|---|---|
| Train SemanticNav on CLIP semantic-goal embeddings rather than ZER's ImageNav-then-language-grounding approach | ZSON with ResNet-9 and Gibson training · ObjectNav (Gibson) | SR: 11.3% -> 15.3% | Improved |
| Use a learned navigation policy instead of frontier-based heuristic exploration | ZSON vs CoW · ObjectNav (MP3D) | SR: 11.1% -> 15.3% | Improved |
| Use a pretrained visual observation encoder (ablation) | ZSON · ObjectNav (Gibson), Gibson-trained model | SR: +9.4% | Improved |
| Use a pretrained visual observation encoder (ablation) | ZSON · ObjectNav (Gibson), HM3D-trained model | SR: +10.4% | Improved |
| Use a pretrained visual observation encoder (ablation) | ZSON · ImageNav (Gibson), Gibson-trained model | SR: +4.5% | Improved |
| Use a pretrained visual observation encoder (ablation) | ZSON · ImageNav (Gibson), HM3D-trained model | SR: +5.8% | Improved |
| Train in 800 HM3D environments instead of 72 Gibson environments, keeping the pretrained encoder (ablation) | ZSON · ObjectNav (Gibson) | SR: +6.6% | Improved |
| Train in 800 HM3D environments instead of 72 Gibson environments, keeping the pretrained encoder (ablation) | ZSON · ImageNav (Gibson) | SR: -0.9% | Hurt |
| Replace scratch ResNet-9 with scratch ResNet-50 (ablation) | ZSON trained on HM3D · ImageNav (Gibson) | SR: 31.1% -> 28.2% | Hurt |
| Replace scratch ResNet-9 with scratch ResNet-50 (ablation) | ZSON trained on HM3D · ObjectNav (Gibson) | SR: 20.9% -> 16.6% | Hurt |
| Finetune with ObjectNav annotations for 25M steps (ablation) | ZSON · ObjectNav (MP3D) | SR: 15.3% -> 22.9% | Improved |
| Finetune with ObjectNav annotations for 25M steps (ablation) | ZSON · ObjectNav (MP3D) | SPL: 4.8% -> 9.2% | Improved |
Later papers that compared with this method
4 later accepted papers report ZSON as a baseline in their results tables; 4 of them report a better number on most of the results they share with it.
- PanoNav: Mapless Zero-Shot Object Navigation with Panoramic Scene Parsing and Dynamic Memory (AAAI 2026 · better on 2 of 2 shared results)
- FloVerse: Floor Plan-Guided Multi-Modal Navigation (CVPR 2026 · better on 1 of 1 shared result)
- Prioritized Semantic Learning for Zero-shot Instance Navigation (ECCV 2024 · better on 4 of 4 shared results)
- ESC: Exploration with Soft Commonsense Constraints for Zero-shot Object Navigation (ICML 2023 · better on 2 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 0.89)
- VoroNav: Voronoi-based Zero-shot Object Navigation with Large Language Model (ICML 2024 · similarity 0.88)
- UniGoal: Towards Universal Zero-shot Goal-oriented Navigation (CVPR 2025 · similarity 0.87)
- Distilling LLM Prior to Flow Model for Generalizable Agent’s Imagination in Object Goal Navigation (NeurIPS 2025 · similarity 0.87)
- RAO-Nav: Probing Omni-Language Models for Zero-shot Semantic Audio-Visual Navigation (NeurIPS 2026 · similarity 0.87)
- GAMap: Zero-Shot Object Goal Navigation with Multi-Scale Geometric-Affordance Guidance (NeurIPS 2024 · similarity 0.87)
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.
