Research map / Probabilistic and scientific modeling / Dynamical systems
The Hippocampal Place Field Gradient: A Bio-inspired Framework Building Multiscale Representation for Better Sample Efficiency
ZHOU Shujun, Junrong Qi, Guozhang Chen
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 |
|---|---|---|---|
| replace RoPE with HIPE | Transformer · One-to-One associative recall | accuracy (%): 81.56 -> 88.53 | Improved |
| replace RoPE with HIPE | Transformer · Block retrieval | accuracy (%): 96.32 -> 97.10 | Improved |
| use HIPE bipartition instead of RoPE | 20M Transformer · WikiText-103 L=2048 | perplexity: 133.81 -> 132.97 | Improved |
| use HIPE bipartition instead of RoPE | 60M Transformer · C4 L=1024 | perplexity: 141.65 -> 130.47 | Improved |
| use HIPE instead of RoPE at larger scale | 300M Transformer · C4 1B tokens | perplexity: 28.93 -> 28.97 | No clear effect |
| use HIPE instead of RoPE with few labels | 300M LoRA · SST-2 100 samples | accuracy (%): 67.3 -> 68.5 | Improved |
| use HIPE instead of RoPE with full training set | 300M LoRA · SST-2 full data | accuracy (%): 82.03 -> 81.77 | Hurt |
| add HIPE (sigma_p=500) to deep layers with local-global attention (ablation) | Local-Global · C4 60M L=1024 | perplexity: 59.36 -> 59.01 | Improved |
| add HIPE to deep layers of global-attention baseline (ablation) | Baseline · C4 60M L=1024 | perplexity: 62.82 -> 62.71 | Improved |
| use uniform sigma_p=50 across all layers instead of bipartition (ablation) | HIPE · WikiText-103 20M L=2048 | perplexity: 133.81 -> 138.45 | Hurt |
| use reverse schedule (broad shallow, sharp deep) instead of sharp-to-broad (ablation) | HIPE · WikiText-103 20M L=2048 | perplexity: 136.22 vs 132.97 | Hurt |
| use continuous gradient instead of discrete bipartite schedule (ablation) | HIPE · WikiText-103 20M L=2048 | perplexity: +2.56 | Hurt |
Closest papers by meaning
- Place Cells as Multi-Scale Position Embeddings: Random Walk Transition Kernels for Path Planning (NeurIPS 2025 · similarity 0.90)
- Learning Place Cell Representations and Context-Dependent Remapping (NeurIPS 2024 · similarity 0.89)
- Trading Place for Space: Increasing Location Resolution Reduces Contextual Capacity in Hippocampal Codes (NeurIPS 2024 · similarity 0.89)
- Time Makes Space: Emergence of Place Fields in Networks Encoding Temporally Continuous Sensory Experiences (NeurIPS 2024 · similarity 0.87)
- Binding in hippocampal-entorhinal circuits enables compositionality in cognitive maps (NeurIPS 2024 · similarity 0.87)
- A Multi-Region Brain Model to Elucidate the Role of Hippocampus in Spatially Embedded Decision-Making (ICML 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.
