Research map / Probabilistic and scientific modeling / Dynamical systems
Emergence of Spatial Representation in an Actor-Critic Agent with Hippocampus-Inspired Sequence Generator
Xiao-Xiong Lin, Yuk-Hoi Yiu, Christian Leibold
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 CA3 sequence generator instead of LSTM core | actor-critic agent · Maze navigation, sparse input | success rate: 0.56 -> 0.86 | Improved |
| use CA3 sequence generator instead of LSTM core under dense input | actor-critic agent · Maze navigation, dense input | success rate: 0.93 -> 0.71 | Hurt |
| use CA3 sequence generator instead of HiPPO-LegS | actor-critic agent · Maze navigation, sparse input | success rate: 0.52 -> 0.86 | Improved |
| shorten sequence length (L=1, no sequences) (ablation) | CA3 agent · Maze navigation | success rate | Hurt |
| increase repetition parameter R from 1 to 8-16 (ablation) | CA3 agent · Maze navigation, frameskip 8 | success rate at 150M frames: 0.788 -> 0.886 (R=12) | Improved |
| increase R to 32 (ablation) | CA3 agent · Maze navigation, frameskip 4 | success rate at 150M frames: 0.902 (R=12) -> 0.758 | Hurt |
| add pixel noise to suppress weak signals (ablation) | DG+CA3 vs dense LSTM · Maze navigation | success rate at 100M frames: DG+CA3 0.693 -> 0.477; LSTM 0.638 -> 0.373 at noise 80 | Mixed |
| permute weights of highest spatial-information decoder units (ablation) | CA3 agent · Maze navigation | trajectory length (frames): 1065 -> 2794 | Hurt |
| permute weights of lowest spatial-information decoder units (ablation) | CA3 agent · Maze navigation | success rate: 100%, 1065 frames | No clear effect |
Closest papers by meaning
- Place Cells as Multi-Scale Position Embeddings: Random Walk Transition Kernels for Path Planning (NeurIPS 2025 · similarity 0.89)
- Time Makes Space: Emergence of Place Fields in Networks Encoding Temporally Continuous Sensory Experiences (NeurIPS 2024 · similarity 0.89)
- REMI: Reconstructing Episodic Memory During Internally Driven Path Planning (NeurIPS 2025 · similarity 0.89)
- A Multi-Region Brain Model to Elucidate the Role of Hippocampus in Spatially Embedded Decision-Making (ICML 2025 · similarity 0.88)
- Binding in hippocampal-entorhinal circuits enables compositionality in cognitive maps (NeurIPS 2024 · similarity 0.88)
- A simple model of co-emergence of grid and place fields (NeurIPS 2026 · similarity 0.88)
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.
