Research map / Architectures and efficiency / RNNs and normalization
A Unified, Scalable Framework for Neural Population Decoding
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 |
|---|---|---|---|
| Pretrain jointly across 100 sessions instead of training separate single-session models | POYO-mp · Center-out tasks, 100-session average | R²: 0.9347 -> 0.9512 | Helped |
| Pretrain jointly across 100 sessions instead of training separate single-session models | POYO-mp · Random target tasks, 100-session average | R²: 0.8402 -> 0.8738 | Helped |
| Fine-tune model weights after unit identification | POYO-mp · Monkey T - CO | R²: 0.9012 -> 0.9379 | Helped |
| Fine-tune model weights after unit identification | POYO-mp · Monkey T - RT | R²: 0.7759 -> 0.8105 | Helped |
| Train across more diverse labs, animals, and tasks | POYO-1 unit identification · NLB-Maze | R²: 0.8962 -> 0.9329 | Helped |
| Use POYO-1 instead of POYO-mp | Pretrained model · NLB-RTT, fine-tuned | R²: 0.7318 -> 0.7378 | Helped |
| Use POYO-1 instead of POYO-mp | Pretrained model · Random target sessions from the same lab | R²: 0.8788 -> 0.8664 | Hurt |
| Increase model depth | Multi-session POYO · Center-out and random target tasks | R² | Helped |
| Increase joint training data from single-session data to all random-target sessions | Multi-session POYO, L=6 · Random target tasks | R²: average 1.2M tokens -> 26M tokens | Helped |
| Add value-rotation to attention (ablation) | POYO attention layers · Neural population decoding | Decoding performance | Helped |
| Rotate only half of the attention head dimensions (ablation) | POYO rotary position encoding · Neural population decoding | Decoding performance | Helped |
| Increase batch size for single-session training (ablation) | POYO single-session models · Single-session decoding | Decoding performance | No clear effect |
Closest papers by meaning
- Towards a "Universal Translator" for Neural Dynamics at Single-Cell, Single-Spike Resolution (NeurIPS 2024 · similarity 0.90)
- Population Transformer: Learning Population-level Representations of Neural Activity (ICLR 2025 · similarity 0.89)
- Neural Encoding and Decoding at Scale (ICML 2025 · similarity 0.89)
- BLEND: Behavior-guided Neural Population Dynamics Modeling via Privileged Knowledge Distillation (ICLR 2025 · similarity 0.88)
- A Scalable, Causal, and Energy Efficient Framework for Neural Decoding with Spiking Neural Networks (NeurIPS 2025 · similarity 0.88)
- Leveraging unlabelled data for generalizable neural population decoding (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.