Research map / Architectures and efficiency / RNNs and normalization
Is a Modular Architecture Enough?
Sarthak Mittal, Yoshua Bengio, Guillaume Lajoie
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 |
|---|---|---|---|
| add perfect rule-based module selection (GT-Modular oracle) to a modular system (ablation) | Modular/Monolithic systems · rule-based synthetic data (MLP, MHA, RNN) | in- and out-of-distribution error/loss | Improved |
| replace Monolithic with end-to-end trained Modular (ablation) | Monolithic · rule-based synthetic data | error/loss | Improved |
| use modularity with few rules (e.g. 2) vs many rules (ablation) | Modular vs Monolithic · rule-based synthetic data | error/loss | Mixed |
| make activation depend only on rule context (Modular-op) instead of (x, c) (ablation) | Modular · rule-based synthetic data | collapse and specialization | Improved |
| increase number of rules from 2 to 32 (ablation) | Modular and Modular-op · rule-based synthetic data | collapse and specialization | Hurt |
| better specialization on training speed (ablation) | GT-Modular · rule-based synthetic data | training convergence | Improved |
Closest papers by meaning
- Breaking Neural Network Scaling Laws with Modularity (ICLR 2025 · similarity 0.88)
- Neural networks are more modular than single neurons suggest (NeurIPS 2026 · similarity 0.86)
- Scalable Optimization in the Modular Norm (NeurIPS 2024 · similarity 0.86)
- Range, not Independence, Drives Modularity in Biologically Inspired Representations (ICLR 2025 · similarity 0.85)
- Learning to grok: Emergence of in-context learning and skill composition in modular arithmetic tasks (NeurIPS 2024 · similarity 0.85)
- Explainable Neural Networks with Guarantee: A Sparse Estimation Approach (AAAI 2025 · similarity 0.84)
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.
