How Finite Brains Generalize Without Bound: Spatial Memory as a Universal Computer
cs.AI
Submitted: 2026-11-05
Updated: 2026-11-16
Terminology
Sources
- Estimating or Propagating Gradients Through Stochastic Neurons for Conditional Computation
- Geometric Deep Learning: Grids, Groups, Graphs, Geodesics, and Gauges
- Extending Context Window of Large Language Models via Positional Interpolation
- Improving the Neural GPU Architecture for Algorithm Learning
- Adaptive Computation Time for Recurrent Neural Networks
- Neural Turing Machines
- Copying Under Repetition: From Prefix Matching to Position Hashing
- When Models Manipulate Manifolds: The Geometry of a Counting Task
- Length Generalization in Arithmetic Transformers
- Language Models Use Trigonometry to Do Addition
- Stabilizing Extrapolation in Looped Transformers via Learned Stochastic Stopping
- Grokking: Generalization Beyond Overfitting on Small Algorithmic Datasets
- Extensions and Limitations of the Neural GPU
- Positional Description Matters for Transformers Arithmetic
- On the Spatiotemporal Dynamics of Generalization in Neural Networks
- What Algorithms can Transformers Learn? A Study in Length Generalization
- Transformers Can Achieve Length Generalization But Not Robustly
Related papers
- Multi-Objective Structured Pruning of LLMs for Latency and Model Size Optimization
- MAVEN-T: Reinforced Heterogeneous Distillation for Real-Time Multi-Agent Trajectory Prediction
- Model Discovery Agent: LLM-assisted Bayesian experiment design for data-efficient discovery of mechanistic world models
- Embeddings for Preferences, Not Semantics
- Can We Trust Item Response Theory for AI Evaluation?
- Diagnosing Task Insensitivity in Language Agents