Research map / Probabilistic and scientific modeling / Variational and generative sampling
NeoNav: Improving the Generalization of Visual Navigation via Generating Next Expected Observations
Qiaoyun Wu, Dinesh Manocha, Jun Wang, Kai Xu
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 four-view instead of front-view only input (ablation) | NeoNav · AVD novel scenes, without stop action | success rate (%): 37.1 -> 47.7 | Improved |
| generate next expected observation via variational sampling (vs non-generative) (ablation) | NeoNav · AVD novel scenes, without stop action | success rate (%): 34.8 -> 47.7 | Improved |
| use mixture-of-posteriors prior instead of standard normal (ablation) | NeoNav · AVD novel scenes, without stop action | success rate (%): 40.3 -> 47.7 | Improved |
| use depth input instead of RGB (ablation) | NeoNav · AVD novel scenes, without stop action | success rate (%): 37.1 -> 47.7 | Improved |
| use depth input instead of RGBD (ablation) | NeoNav · AVD novel scenes, without stop action | success rate (%): 42.2 -> 47.7 | Improved |
| NeoNav vs best baseline with stop action | NeoNav · AVD novel scenes, with stop action | success rate (%): 14.2 -> 19.9 | Improved |
| NeoNav vs best baseline in AI2-THOR | NeoNav · AI2-THOR novel scenes, with stop action | success rate (%): 11.7 -> 16.7 | Improved |
| NeoNav vs TD-A3C in bathroom scenes | NeoNav · AI2-THOR bathrooms | success rate (%): 24.3 -> 21.9 | Hurt |
| fewer training scenes (8 to 2) with NeoNav | NeoNav · AVD without stop action | success rate (%): 47.9 -> 35.1 | Hurt |
| NeoNav vs TD-Semantic with depth input in semantic-driven navigation | NeoNav · AVD semantic-driven navigation | success rate (%): 53.6 -> 66.3 | Improved |
Closest papers by meaning
- Scaling Data Generation in Vision-and-Language Navigation (ICCV 2023 · similarity 0.90)
- LatentPilot: Scene-Aware Vision-and-Language Navigation by Dreaming Ahead with Latent Visual Reasoning (ECCV 2026 · similarity 0.90)
- NavGPT: Explicit Reasoning in Vision-and-Language Navigation with Large Language Models (AAAI 2024 · similarity 0.88)
- Seeing through Uncertainty: Robust Task-Oriented Optimization in Visual Navigation (NeurIPS 2025 · similarity 0.88)
- Image2Sim: Scaling Embodied Navigation via Generative Neural Simulator (NeurIPS 2026 · similarity 0.87)
- Navigation World Models (CVPR 2025 · 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.
