Research map / Label-efficient and robust learning / Recognition and long-tail
Learning Structure from the Ground up---Hierarchical Representation Learning by Chunking
Shuchen Wu, Noemi Elteto, Ishita Dasgupta, Eric Schulz
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 HCM instead of an RNN to model human reaction times | HCM · Human reaction time regression (47 participants) | regression coefficient beta: 9.24 -> 16.74 | Improved |
| use HCM instead of an RNN to model human reaction times | HCM · Human reaction time regression (47 participants) | Kendall tau: 0.085 -> 0.165 | Improved |
| use HCM instead of an RNN to model human reaction times | HCM · Human reaction time regression (47 participants) | BIC: 314236.4 -> 313586.5 | Improved |
| use HCM instead of a 3-layer RNN as the learner on hierarchical generative sequences | HCM · Synthetic hierarchical sequences, depths d | KL divergence convergence speed | Improved |
| increase depth of the generative hierarchy | HCM versus RNN · Synthetic hierarchical sequences | data-efficiency advantage over RNN | Improved |
| pre-train HCM on an interfering environment before switching | HCM · Environment transfer | early performance over first 500 trials | Hurt |
| pre-train HCM on a facilitative environment before switching | HCM · Environment transfer | early performance over first 500 trials | Improved |
| increase participant age | HCM · Resting-state fMRI | average chunk size: r = 0.23 | Improved |
Closest papers by meaning
- Concept-Guided Interpretability via Neural Chunking (NeurIPS 2025 · similarity 0.88)
- Building, Reusing, and Generalizing Abstract Representations from Concrete Sequences (ICLR 2025 · similarity 0.87)
- Flexible Context-Driven Sensory Processing in Dynamical Vision Models (NeurIPS 2024 · similarity 0.85)
- Learning Discrete Concepts in Latent Hierarchical Models (NeurIPS 2024 · similarity 0.85)
- Large Language Models as Model Organisms for Human Associative Learning (NeurIPS 2025 · similarity 0.85)
- Neural networks are more modular than single neurons suggest (NeurIPS 2026 · similarity 0.85)
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.
