Research map / Detection and segmentation / Open-set and few-shot detection
Function-centric Bayesian Network for Zero-Shot Object Goal Navigation
Sixian Zhang, Xinyao Yu, Xinhang Song, Yiyao Wang, Shuqiang Jiang
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 |
|---|---|---|---|
| use CounterfactCoT instead of direct prompting for edge estimation (ablation) | FBN-Llama 7B · HM3D val | SR/SPL: 45.5 -> 53.2 SR, 21.6 -> 30.7 SPL | Improved |
| use CounterfactCoT instead of standard CoT (ablation) | FBN-Llama 7B · HM3D val | SR/SPL: 49.8 -> 53.2 SR, 26.5 -> 30.7 SPL | Improved |
| use CounterfactCoT instead of direct prompting (ablation) | FBN-GPT 4o · HM3D val | SR/SPL: 50.1 -> 58.8 SR, 24.7 -> 31.2 SPL | Improved |
| use CounterfactCoT instead of standard CoT (ablation) | FBN-GPT 4o · HM3D val | SR/SPL: 56.0 -> 58.8 SR, 27.5 -> 31.2 SPL | Improved |
| use CoT instead of direct prompting (ablation) | FBN-Llama 7B · HM3D val | SR/SPL: 45.5 -> 49.8 SR, 21.6 -> 26.5 SPL | Improved |
| replace Llama 7B with GPT-4o as the LLM | FBN with CounterfactCoT · HM3D val | SR/SPL: 53.2 -> 58.8 SR, 30.7 -> 31.2 SPL | Improved |
| add accurate scene recognition in the semantic map (vs random noise) (ablation) | FBN-Llama · HM3D val | SR/SPL: 50.1 -> 53.2 SR, 28.9 -> 30.7 SPL | Improved |
| add accurate object recognition in the semantic map (vs random noise) (ablation) | FBN-Llama · HM3D val | SR/SPL: 48.3 -> 53.2 SR, 25.0 -> 30.7 SPL | Improved |
| add accurate scene and object recognition (vs noise in both) (ablation) | FBN-Llama · HM3D val | SR/SPL: 41.7 -> 53.2 SR, 21.8 -> 30.7 SPL | Improved |
| add object-to-function edges to scene-to-function only (ablation) | FBN-Llama · HM3D val | SR/SPL: 48.6 -> 53.2 SR, 26.3 -> 30.7 SPL | Improved |
| add scene-to-function edges to object-to-function only (ablation) | FBN-Llama · HM3D val | SR/SPL: 51.8 -> 53.2 SR, 29.1 -> 30.7 SPL | Improved |
| FBN (Llama) versus training-free SGM* | FBN-Llama · MP3D val | SR: 30.1 -> 41.1 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- SG-Nav — SG-Nav: Online 3D Scene Graph Prompting for LLM-based Zero-shot Object Navigation (NeurIPS 2024) · better on 5 of 5 shared results
- GAMap — GAMap: Zero-Shot Object Goal Navigation with Multi-Scale Geometric-Affordance Guidance (NeurIPS 2024) · 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
- ZSON: Zero-Shot Object-Goal Navigation using Multimodal Goal Embeddings (NeurIPS 2022 · similarity 0.87)
- BeliefMapNav: 3D Voxel-Based Belief Map for Zero-Shot Object Navigation (NeurIPS 2025 · similarity 0.87)
- UniGoal: Towards Universal Zero-shot Goal-oriented Navigation (CVPR 2025 · similarity 0.86)
- SG-Nav: Online 3D Scene Graph Prompting for LLM-based Zero-shot Object Navigation (NeurIPS 2024 · similarity 0.86)
- GAMap: Zero-Shot Object Goal Navigation with Multi-Scale Geometric-Affordance Guidance (NeurIPS 2024 · similarity 0.85)
- Distilling LLM Prior to Flow Model for Generalizable Agent’s Imagination in Object Goal Navigation (NeurIPS 2025 · similarity 0.84)
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.
