Research map / Multi-agent systems and games / Human-AI and goal-conditioned
Semi-parametric topological memory for navigation
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 SPTM rather than the strongest reported baseline | goal-directed navigation with walkthrough memory · Test mazes | success rate (%): 100 versus best baseline values 53, 39, 45, 23, 32, 36, 22 | Helped |
| add visual shortcut edges to the graph (ablation) | SPTM · Validation mazes | success rate (%): 85 -> 100; 55 -> 98; 50 -> 100 | Helped |
| use temporally consistent localization instead of independent per-frame localization (ablation) | SPTM · Val-3 | success rate (%): 52 -> 100 | Helped |
| use temporally consistent localization instead of independent per-frame localization (ablation) | SPTM · Val-1 and Val-2 | success rate (%): 95 -> 100; 87 -> 98 | Helped |
| replace the retrieval network with normalized per-pixel matching (ablation) | SPTM · Validation mazes | success rate (%): 100 -> 6; 98 -> 2; 100 -> 1 | Hurt |
| replace the retrieval network with unnormalized per-pixel matching (ablation) | SPTM · Validation mazes | success rate (%): 100 -> 10; 98 -> 8; 100 -> 7 | Hurt |
| replace the locomotion network with recorded exploration actions (teach-and-repeat) (ablation) | SPTM · Validation mazes | success rate (%): 100 -> 45; 98 -> 36; 100 -> 30 | Hurt |
| use homogeneous textures (ablation) | SPTM · Val-3 | success rate (%): 100 -> 55 | Hurt |
| use homogeneous textures (ablation) | SPTM · Test-1, Test-4, and Test-5 | success rate (%): 100 -> 98; 100 -> 76; 100 -> 75 | Hurt |
| use automated rather than human exploration footage (ablation) | SPTM · Test-1, Test-4, and Test-5 | success rate (%): 100 -> 94; 100 -> 93; 100 -> 91 | Hurt |
| vary the number of graph shortcuts from 2000 to 1000 or 4000 (ablation) | SPTM · Validation mazes | success rate (%): 100, 98, 100; 99, 92, 90; 100, 98, 100 | No clear effect |
| subsample the memory by factor 1 or 8 instead of 4 (ablation) | SPTM · Validation mazes | success rate (%): 100, 98, 100; 76, 47, 73; 66, 92, 66 | Mixed |
Closest papers by meaning
- Neural Topological SLAM for Visual Navigation (CVPR 2020 · similarity 0.88)
- Emergence of Maps in the Memories of Blind Navigation Agents (ICLR 2023 · similarity 0.87)
- MemoNav: Working Memory Model for Visual Navigation (CVPR 2024 · similarity 0.87)
- Hierarchical Semantic-Augmented Navigation: Optimal Transport and Graph-Driven Reasoning for Vision-Language Navigation (NeurIPS 2025 · similarity 0.85)
- Reasoning in Visual Navigation of End-to-end Trained Agents: A Dynamical Systems Approach (CVPR 2025 · similarity 0.85)
- Learning with a Mole: Transferable latent spatial representations for navigation without reconstruction (ICLR 2024 · 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.