Research map / Architectures and efficiency / RNNs and normalization
Visual Memory for Robust Path Following
Ashish Kumar, Saurabh Gupta, David Fouhey, Sergey Levine, Jitendra Malik
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 RPF instead of open-loop replay | path following · SBPD area4 | Success Rate: 0.216 -> 0.866 | Improved |
| Use RPF instead of 3D reconstruction and localization | path following · SBPD area4 | Success Rate: 0.826 -> 0.866 | Improved |
| Add visual features to the path memory (ablation) | RPF · SBPD area4, Following | Success Rate: 0.782 -> 0.866 | Improved |
| Add visual features to the path memory (ablation) | RPF · SBPD area4, Homing | Success Rate: 0.780 -> 0.866 | Improved |
| Use RPF instead of 3D reconstruction and localization | homing · SBPD area4 | Success Rate: 0.000 -> 0.866 | Improved |
| Replace the recurrent controller with a feed-forward network (ablation) | RPF · path retracing | Success Rate | Hurt |
| Replace the learned pointer increment with a constant increment (ablation) | RPF · path retracing | Success Rate | Hurt |
| Use a standard GRU without the proposed sequential memory (ablation) | path retracing · path retracing | Success Rate | Hurt |
| Use trained nearest-neighbor action voting | path retracing · path retracing | Success Rate | Hurt |
| Increase path length beyond the training length | RPF · novel environments | Success Rate: 3× | Mixed |
Closest papers by meaning
- Grounding Driving VLA via Inverse Kinematics (NeurIPS 2026 · similarity 0.86)
- VLMs Trace Without Tracking: Diagnosing Failures in Visual Path Following (NeurIPS 2026 · similarity 0.85)
- C-NAV: Towards Self-Evolving Continual Object Navigation in Open World (NeurIPS 2025 · similarity 0.84)
- Learning with a Mole: Transferable latent spatial representations for navigation without reconstruction (ICLR 2024 · similarity 0.84)
- Navigation World Models (CVPR 2025 · similarity 0.84)
- Affordances-Oriented Planning Using Foundation Models for Continuous Vision-Language Navigation (AAAI 2025 · similarity 0.83)
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.
